Final Report on New Jersey’s Empirical

Study of Jury Selection Practices and

Jury Representativeness

Prepared for the New Jersey Supreme Court

Mary R. Rose, Ph.D.*

June 1, 2021

*The author gratefully acknowledges Brian McLaughlin, Lisa Burke, and Jessica

Lewis-Kelly of the Administrative Office of the Courts for their consistent

professionalism, incredible dedication to this project, and the supportive assistance

they provided to me.

i

AUTHOR’S PREFACE

In this report, I use data from 95 trials across 14 counties to examine the

representativeness of juries in New Jersey and to explore potential sources of

attrition from jury service that are systemic (system-wide) and/or systematic

(attributable to processes). Significantly, and surprisingly, the profile of people

selected for juries in most – but not all – New Jersey counties reasonably

represented the profile of people who appear at court (there is, however,

systemic minority attrition among the group who appears at courthouses).

Precisely because results were surprising, I use this preface to ensure clarity on

which conclusions can and cannot be drawn from these data.

Supported Conclusions

As outlined in more detail in the report, the data support the following

conclusions:

*The data support the conclusion that people who appear for jury service

do not fully represent their communities.

*The report finds that, sporadically, final juries fail to represent the panels

from which they were drawn, and in some instances, all members of a minority

group failed to be seated on a jury. The latter was particularly likely when there

were few minority-group members in the venire and in civil cases.

ii

*The data support the conclusion that criminal juries do a better job

representing communities than do civil juries, likely because criminal juries are

larger in size (12 vs. 6 jurors).

*The data support the conclusion that the challenge for cause is the most

common way for people to be removed from jury service in criminal cases, and it

is the second-leading source of attrition from civil cases. The data do not suggest

that attorneys must routinely use their peremptory challenges to compensate for

judges who are highly conservative about granting challenges for cause.

*The report finds that New Jersey forms large panels for jury selection in

routine trials, larger than is typical in most other areas of the country. High

proportions of jurors are “not used” (i.e., neither dismissed through a challenge

for cause or peremptory challenge, nor selected), and this is especially true in civil

trials.

*The report finds that attorneys almost never use their full complement of

peremptory challenges.

*The report finds that although peremptory challenges can be linked

sporadically to minority-group attrition patterns, peremptory challenges are not

the primary reason why African American, Latino, or Asian jurors fail to make it on

to a jury.

iii

Unsupported/Unwarranted Conclusions

At the same time, these data CANNOT be used to conclude any of the

following:

*The report does NOT conclude that New Jersey courts should be

unconcerned about racial or ethnic underrepresentation in jury selection. In

contrast, the findings indicate that the processes that determine who appears at

the courthouse constitute a systemic source of minority-group attrition because

concerning levels underrepresentation appeared in nearly all areas studied, a

finding consistent with many studies of other areas and courts.i Determining

which systematic processes account for this common pattern is beyond the scope

of this report, but nothing in this report should suggest that New Jersey need not

consider this important question.

Excepting that the size of civil juries likely impedes better representation,

most practices occurring at the courthouse did not produce systemic race- or

ethnicity-based attrition. There were either no relationships between

i See, e.g., Jacinta M. Gau, A Jury of Whose Peers? The Impact of Selection

Procedures on Racial Composition and the Prevalence of Majority-White Juries, 39

JOURNAL OF CRIME AND JUSTICE 75 (2016); Mary R. Rose, Raul S. Casarez, and Carmen

M. Gutierrez, Jury Representation in the Modern Era: Evidence from Federal

Courts, 15 JOURNAL OF EMPIRICAL LEGAL STUDIES 378 (2018).

iv

race/ethnicity and selection processes, or instances of underrepresentation were

sporadic and county-specific. Significantly, whether these sporadic instances of

underrepresentation reflect recurrent, systematic problems in those specific

counties cannot be determined from these data. These data represent a snapshot

of one, roughly six-week point in time, and some areas contributed only a few

trials to the dataset in that time.

Rather than being dismissed or minimized, the instances of

underrepresentation described in this report offer an opportunity for counties to

consider why minority groups failed to make it onto some juries at all or did so in

lower proportions than their representation in the venire. Consistent with their

constitutional obligations, courts in all areas, even those with little minority

attrition in these data, should continue to look for concerning patterns of

underrepresentation in their jury pools and juries.

*The data do NOT support a conclusion that the number of peremptory

challenges allocated to attorneys does no harm to jury selection practices and

outcomes. Compared to other states, New Jersey has an unusually high number of

peremptory challenges allotted to attorneys; it also has unusually large panels

called up for jury selection in a given case. As suggested in Section VII of this

report, conceivably these two factors are related because panel sizes may need to

v

be large to accommodate the possibility of both sides using all of their strikes. If

true, this would constitute one way that peremptory challenge practices

negatively affect other aspects of jury selection. New Jersey courts should rightly

be concerned if prospective jurors feel that their time is wasted when they are

“not used,” an outcome the data reveals to be commonplace. Nonetheless, I did

not have access to the type of data (e.g., interviews with jury clerks) that would

confirm the connection between panel size planning and the expected exercise of

peremptory challenges, nor did I speak to New Jersey jurors about their reactions

to their experiences.

More importantly, the ability to observe a relationship between attorneys’

use of large numbers of peremptory challenges and levels of minority

representation was limited. This is because attorneys almost never used their full

allotment of challenges. In criminal cases, prosecutors used about four challenges

on average, and defendants used about six, far below the large number allotted.

Studies in other areas, such as capital cases in other jurisdictions, suggest that

using large numbers of peremptory challenges can negatively affect

representation and have other negative effects. For example, counsel in capital

cases often have a sizeable number of challenges. According to one study of

Philadelphia, in more than half of the cases in which young African American men

vi

appeared in the venire for capital cases, prosecutors were able to remove all of

them.ii They were able to remove all young Black women in 41% of the cases that

contained at least one person from this group.iii Because prosecutors in criminal

cases are disproportionately likely to use peremptory challenges on minority-

group members, particularly African Americans, they hold a structural advantage

in the process. As researchers in the Philadelphia study concluded: “The

Commonwealth was more successful than defense counsel at eliminating its

prime targets [young Black men, young Black women, and middle-aged Black

women] from jury service, an outcome that reflects the different size pools of

each side’s prime target groups.”iv

Beyond representativeness, aggressive use of peremptory challenges also

risks harm to jurors’ perceptions of the jury selection process. In a rare study of

post-jury selection interviews with over 100 people excused through the

peremptory challenge, I found that prospective jurors who said that they were

ii E.g., Baldus, David C., George Woodworth, David Zuckerman, and Neil Alan

Weiner. The Use of Peremptory Challenges in Capital Murder Trials: A Legal and

Empirical Analysis, 3 UNIVERSITY OF PENNSYLVANIA JOURNAL OF CONSTITUTIONAL LAW 3, 98

(2001).

iii Id at 99.

iv Id at 98.

vii

excused because of some personal characteristic (e.g., gender, age, race,

occupation) were more likely to view the decision as unfair and to believe that the

attorney did not accurately assess their abilities to be impartial, compared to

people who believed they were struck for other reasons.v Although most people

excused from service are accepting – and sometimes relieved – at the outcome,

some prospective jurors can be offended when they have appeared for service

but end up excused with no explanation, particularly if they believe they have

been crassly stereotyped. As one person in my study said: “All they say is ‘See you

next year.’ You won’t see me.”vi The greater the number of challenges used, the

more likely it is that jurors will be confused about the reasons and assume

decisions were based on irrelevant characteristics.

Therefore, in sum, studies from other areas strongly suggest that an

aggressive use of available peremptories in a case can distort representativeness

v Those attributing their dismissal to a personal characteristic were compared to those who suspected they were excused because they had experiences with crime/the legal system, or because they behaved in ways during jury selection that might raise concerns about their abilities to be fair (e.g., they were hesitant in their answers). See Mary R. Rose, A Voir Dire of Voir Dire: Listening to Jurors’ Views Regarding the Peremptory Challenge, 78 CHICAGO-KENT L. REV 1061, 1086 (2003). vi Id at 1095.

viii

and harm perceptions of the process. A key reason that this study cannot

conclude that the current large allotment of peremptory challenges does no harm

to jury selection in New Jersey is simply because this study could not test for this

possibility: the data revealed almost no instance of attorneys using close to their

allotted amount.

*The data do NOT support the conclusion that attorneys ignore race when

using peremptory challenges. Multiple studies of jury selection from other

regions find that attorneys demonstrate adversarial, race-based patterns in how

they exercise peremptory challenges (e.g., the state disproportionately dismisses

African Americans in criminal cases, and the defense disproportionately dismisses

White prospective jurors).vii Consistent with this broader academic literature, this

report finds statistical evidence that defense attorneys in criminal cases were

disproportionately likely to dismiss White jurors rather than African American

jurors. (Other parties in other trial types did not exercise enough peremptory

vii E.g., Baldus et al., supra, note ii; Shari Seidman Diamond, Destiny Peery, Francis J. Dolan, and Emily Dolan, Achieving Diversity on the Jury: Jury Size and Peremptory Challenges, 6 JOURNAL OF EMPIRICAL LEGAL STUDIES 425 (2010); Gau, supra, note i; Catherine M. Grosso and Barbara O'Brien, A Stubborn Legacy: The Overwhelming Importance of Race in Jury Selection in 173 Post-Batson North Carolina Capital Trials, 97 IOWA L. REV. 1531 (2012). Mary R. Rose, The Peremptory Challenge Accused of Race or Gender Discrimination? Some Data from One County, 23 LAW & HUMAN BEHAVIOR 695 (1999).

ix

challenges on minority group members for findings to emerge as statistically

significant.) Further, there were sporadic instances of peremptory challenges’

removing either all members of one minority group or removing 25% or more of a

group, the standard I adopted for “concerning” levels of

underrepresentation/attrition. In short, even though attorney behavior did not

stand out as the key systematic explanation for levels of representation on juries,

evidence for some amount of racial patterning of strikes did emerge.

*The data do NOT support the notion that judges grant “too many”

challenges for cause. As noted in Section VI, a high proportion of jurors,

particularly in criminal cases, exit jury service through the challenge for cause.

Data about challenges for cause in other states and regions are limited; hence, it

is not possible to determine with certainty whether or how New Jersey’s practices

surrounding challenges for cause may be unusual.viii Yet even if New Jersey’s

judges are out of step with other areas, it bears mentioning that scholarly work

viii Results from the few existing studies on how much attrition stems from cause challenges are mixed. In the study by Rose, id, peremptory challenges were the more common means for jurors to be excused; in data from Gau, supra, note i, challenges for cause removed roughly as many people as peremptory challenges.

x

typically finds judges to be overly cautious about removing jurors for cause,ix

aiming to “rehabilitate” people whose ability to be fair presents grave to concerns

to parties or other court observers.x The “right” amount of challenges for cause

can be assessed in a number of different ways, but this report does not conclude

that judges exercise them unnecessarily, nor that they harm racial representation.

****

As this preface and report make clear, this study’s initial analysis answers

some questions while inviting additional work to explore others. Therefore, New

Jersey will likely want to continue its self-study. I strongly encourage New Jersey

to do so, including through additional data it may collect in the future. New Jersey

and the Administrative Office of the Courts are to be commended for

commissioning this critical look at its practices and for inviting and trusting the

perspective of the scholarly community. I hope I and other outside experts can

and will assist the Courts in any future efforts to improve its jury trials.

ix See, e.g., Mary R. Rose, and Shari Seidman Diamond. Judging Bias: Juror Confidence and Judicial Rulings on Challenges for Cause, 42 LAW & SOCIETY REVIEW 513 (2008). x Christopher A. Crocker, Rehabilitation of the Juror Rehabilitation Doctrine, 37 GEORGIA LAW REVIEW 1471 (2003); Caroline B. Crocker, and Margaret Bull Kovera, The Effects of Rehabilitative Voir Dire on Juror Bias and Decision Making, 34 LAW &

HUMAN BEHAVIOR 212 (2010).

2

TABLE OF CONTENTS

Topic Page

Executive Summary 6

Section I. A Study to Observe Race, Latino Ethnicity, and Gender in New Jersey

Courts 21

A. Cleaning data 23

B. Conclusion and Recommendation #1 25

Section II. Response Rates, Determining Usable Data, and Measurement Issues 30

A. Response rates 30

B. Final dataset 33

C. Recommendations #2a., #2b, and #2c regarding future measurement

efforts 34

Section III. County Racial/Ethnic Demographic Profiles 36

A. Absolute and comparative disparities across counties 36

B. Conclusions and Recommendation #3 44

Section IV. Race, Ethnicity, and Serving on a Jury in New Jersey Courts 47

A. Jury composition 47

B. Criminal vs. civil juries 50

C. Conclusions and Recommendations #4 and #5 57

Section V. Peremptory Challenges 60

3

A. Peremptory challenges and minority underrepresentation 60

1. African Americans 61

2. Latinos 65

3. Asian Americans 68

4. Summary 69

B. Peremptory use overall 69

C. Conclusions 74

Section VI. Challenges for Cause 76

A. Cause challenges and peremptory challenges 78

B. Cause challenges and jury diversity 81

C. Conclusions and Recommendation #6 86

Section VII. The Size of Jury Venires 87

A. The “not used” category 87

B. Recommendation #7 93

Appendix A. Handling Missing Questionnaire Responses and Missing

Questionnaires 94

A. Skipping the Hispanic/Latino or the race question 96

B. Missing whole surveys 98

C. Handling multiple answers on race 98

D. Conclusion and Recommendations #2a, #2b, and #2c 102

4

Appendix B. Statistical Testing of Underrepresentation on Juries 105

A. The meaning of a p-value 105

B. Choosing the appropriate statistical test 107

C. Results for predicting jury participation 108

List of Tables (Tables appear in the sections indicated by the Roman

numerals in their titles.)

Table II.1: Response Rates, and Responses to Race/Ethnicity Questions, Across

Counties of Study.

Table III.1: Racial/Ethnic Profile of Counties in Study, American Community

Survey Data, 2014 – 2018, Table B05003 (Adult Native-Born and

Naturalized Citizens).

Table III.2: Racial/Ethnic Profile, Overall and by County, Based on Respondents

to Survey.

Table III.3: Absolute and Comparative Disparities: All People Appearing for

Service Who Filled Out Surveys.

Table IV.1: Comparison of Venire and Jury Composition, For All Juries, by

County.

Table IV.2: Comparison of Venire and Jury Composition, For Criminal Cases,

by County.

Table IV.3: Comparison of Venire and Jury Composition, For Civil Cases, by

County.

Table V.1: Table V.A. Use of Peremptory Challenges on African Americans

Across Fourteen Counties: Means, Standard Deviations, and Maximum

Number Used, for Criminal and Civil Cases.

5

Table V.2: Use of Peremptory Challenges on Latinos Across Fourteen Counties:

Means, Standard Deviations, and Maximum Number Used, for Criminal and

Civil Cases.

Table V.3: Use of Peremptory Challenges on Asian Americans Across Fourteen

Counties: Means, Standard Deviations, and Maximum Number Used, for

Criminal and Civil Cases.

Table V.4: Use of Peremptory Challenges Across Fourteen Counties: Means,

Standard Deviations, and Maximum Number Used, for Criminal and Civil

Cases.

Table VI.1: Voir Dire Outcomes by County and Trial Type.

Table VII.1: Venire size, by County and Overall.

6

EXECUTIVE SUMMARY

Section I. A Study to Observe Race, Latino Ethnicity, and Gender in New

Jersey Courts

This report presents results and recommendations based on a study of the

race, Latino ethnicity, and gender of the people who appeared for jury service

during a several-week period in September and October of 2018 in fourteen

counties in New Jersey, counties that cover 84% of the state’s population.1

Producing the data, which was done by personnel in each courthouse and at the

Administrative Office of the Courts, and analyzing and writing up the results –

done by me, in conjunction with Dr. Marc Musick of the University of Texas, an

expert on large datasets – was incredibly labor-intensive. This massive effort

demonstrates New Jersey’s commitment to meeting both its constitutional

obligations and its own policy values concerning jury representativeness, thus

making New Jersey a leader among states in this arena.

However, given the work it took to produce a one-time study such as this,

and given the important constitutional and policy aims at stake, my first

recommendation to New Jersey is that it develop a system to routinely measure, at

minimum, the race, ethnicity, and gender of all persons appearing for service.

1 The study gathered data from 15 counties. Due to high levels of nonresponse to fielded

questionnaires, one county had to be eliminated from study.

7

Courts should collect information on cognizable groups in a fashion that permits

regular in-house analyses to more quickly and inexpensively check for problems in

patterns of representativeness. Such a system should be appropriate for how New

Jersey summons and interacts with people at the courthouse, should make it simple

for prospective jurors to report their background characteristics to courts, and

allow New Jersey to store and review these data (see Recommendations 2a – 2c for

suggestions on how to best measure race/ethnicity).

Consistent with Supreme Court rulings, parties should have access to de-

identified data to pursue or defend claims about the representativeness of

“cognizable groups” in jury pools. The complexity of this study shows that current

means of obtaining race data are cumbersome, and it is unfair to burden parties

with trying to collect comparable racial or ethnic data themselves. Ideally, a new

system could also make de-identified data available to scholars or other members

of the public who make reasonable requests. Such transparency increases public

trust. But just as importantly, as the findings of this report reveal, there is much

that New Jersey does well in securing racially representative juries. A reliable and

usable system for collecting and analyzing, at minimum, the race, ethnicity, and

gender of all who appear for jury service constitutes the best way to ensure that the

state continues its successes, and it would educate others about the strengths and

weak spots in New Jersey’s system.

8

Recommendation #1: New Jersey courts should develop a means of

asking all people who appear at the courthouse to identify how they

categorize themselves in terms of their race, Latino/a ethnicity, and

gender. This should be incorporated into any existing questionnaires

prospective jurors may fill out, or New Jersey should explore a new

means of getting this information from all individuals. Such data

should be used for in-house analyses of the effect of court practices on

jury representativeness, and de-identified data (or detailed reports

based on the data) should be available to parties pursing claims about

levels of jury representativeness. New Jersey Courts should also

consider making data available to scholars or members of the public

seeking to understand the effect of court practices on representation.

Section II. Response Rates, Determining Usable Data, and Measurement

Issues

In this study, approximately 85% of all people appearing at courthouses

during the study period provided questionnaire data, although this rate varied

across the 14 counties that contributed usable data. Some trials in the dataset had to

be omitted because they commenced before the study period, or they continued

after the study period concluded. Following omission of these cases, or instances in

which survey nonresponse was excessive, the final dataset contained 7,407

observations involving a civil or criminal voir dire/trial experience, together with

the 5,055 people across the 14 counties who were “pool only” (no voir dire

experience); this constitutes the dataset that produced the findings in this report.

Some experienced more than one voir dire. Hence, these 12,462 observations

9

represent 12,263 unique people, of whom 10,358 also had questionnaire data about

race, Latino ethnicity, and gender.

In Appendix A, “Handling Missing Questionnaire Responses and Missing

Questionnaires,” I describe choices I made when analyzing the data, given that

some observations lacked survey data, and given that some people skipped

questions about Latino identity, about race, or, as permitted, they selected multiple

categories in the race question. Should this study be replicated, or, as

Recommendation #1 advocates, should New Jersey design a system to routinely

measure race and ethnicity, Appendix A offers a technical primer on questionnaire

design, providing the bases for a series of recommendations to improve

measurement of race and ethnicity, particularly in light of the law’s demand that

representation concern a “cognizable group.”

Specifically, Recommendations, #2a, #2b, and #2c advise New Jersey,

respectively, to:

(a) avoid encouraging people to treat questionnaire responses as optional in

order to limit the number of people who fail to provide data; to encourage

responses and engender trust, the form can include an explanation of the need to

ask about race/ethnicity or gender;

(b) New Jersey should encourage people to select a single category that best

describes their racial identity, given that representativeness claims depend upon

10

being able to identify a “cognizable group” that is underrepresented; one category

can be “Multiracial”; and

(c) a question about Latino/Hispanic identity should always precede rather

than follow a question about race, as people tend to treat questions about Latino

identity as a second question about race, making non-Latinos disproportionately

likely to skip the Latino question.

Section III. County Racial/Ethnic Demographic Profiles

Consistent with recommendations in the scholarship on jury

representativeness, this study adopts a standard of a 25% or greater comparative

disparity to signal concerning levels of jury representativeness; that is, I note

instances in which at least 25% of a group have experienced attrition across a

given phase of the jury selection process, and I label these “substantial” or

“concerning.”

A comparison of the jury pools in this study to the communities from which

they were drawn reveals that substantial underrepresentation of African Americans

is commonplace. Just four counties having comparative disparities of African

Americans in their pools of less than 25%. Unlike patterns for other racial/ethnic

groups, no county saw even slight overrepresentation of African Americans among

those appearing at court. By contrast, for Latinos, a few specific areas exhibited at

11

least a 25% comparative disparity in their pools, but underrepresentation was not

as consistent as it was for African Americans. Asian Americans were consistently

overrepresented in jury pools in these areas.

Results for African Americans are consistent with other studies of

representation, and are challenging for courts to address. The reasons why people

may not make it to a courthouse after being summoned are multiple and not fully

under courts’ control to change. In addition, greater summons enforcement can be

costly and risk negative consequences associated with increasing levels of law

enforcement and surveillance on communities that may already be alienated from

police and sheriff officers. Nonetheless, if New Jersey’s goal is to create juries that

better represent their communities, then it must better understand and address

sources of attrition in who makes it to the courthouse. Such investigation may

identify the need to reach out to people in some manner to ensure respect for the

courts’ orders to appear for service, either through greater use of reminders about a

service date or some form of enforcement. Changes to procedures at the courthouse

will yield only so much success in improving the representativeness of juries if the

basic disparities between pools and the broader community are not addressed.

Recommendation #3: New Jersey’s next intensive study of jury

representativeness should be aimed at understanding the source(s) of

why jury pools – the groups of people who appear at the courthouse

for service – consistently and substantially underrepresent African

Americans. New Jersey should consider how to improve summons

response in ways that are appropriate for a given community and that

12

actually generate yields in participation, but which minimize costs to

the court and to the underrepresented communities. New Jersey may

need to enhance its system for reminding people about their assigned

service date and explore reasonable methods of summons

enforcement.

Section IV. Race, Ethnicity and Serving on a Jury in New Jersey Courts

The heart of this intensive research effort is the study of the outcomes of the

7,407 people who experienced voir dire during the study period, and whether

people’s race or ethnicity predicted their selection to a jury. The first conclusion

from analyses of minority representation across fourteen counties is a surprising

one: In many ways, New Jersey courts are doing an admirable job ensuring that

minority-groups participate on criminal and civil juries, based on their venire

composition. In the majority of counties, representation levels of African

Americans, Latinos, and Asian Americans on juries resembled or exceeded the

group’s prevalence in the venires, and when all types of trials were combined

together, typically only two or three counties exhibited at least a 25% comparative

disparity. Underrepresentation of Asian Americans was more commonplace across

counties, but several of these instances reflect the low proportions of this group in

venires. In analyses described and reported in Appendix B, New Jersey courts do

not produce any statistically significant findings that non-Hispanic Whites are

more likely to appear on juries than other groups (although Appendix B explains

why this is a very conservative test of underrepresentation).

13

In analyses that separated results by trial type, criminal versus civil, I found

that, typically, groups were not underrepresented on both civil and criminal cases.

Just one county underrepresented African Americans on both criminal and civil

cases, and a different county had concerning levels of underrepresentation for

Latinos on both civil and criminal cases. I observed little evidence that African

Americans are particularly likely to be underrepresented on criminal cases. By

contrast, multiple indicators suggest that civil cases are more likely to

underrepresent groups compared to criminal cases. In two counties, for example,

despite non-trivial proportions of African Americans in their venires, African

Americans were overrepresented on each county’s lone criminal trial, whereas

multiple civil cases seated no African Americans. In addition, civil juries were

more likely than larger-sized criminal juries to fail to have any African American

or Latino members.

Following these analyses, I make the following recommendations:

Recommendation #4: New Jersey courts should continue their

analysis of the practices that contribute to representative juries being

drawn from venires. Precisely because the juries in these fourteen

areas commonly profiled well on indicators of representativeness,

New Jersey should better understand what exactly contributes to this

result in multiple counties. As the study also points to particular areas

that have more consistent underrepresentation patterns, either across

case type or within case type, these areas should conduct a self-study

to assess what may be happening in these cases. (Sections V, VI, and

VII begin some of this work.) This is particularly important in the

areas in which no African Americans appeared on the mass of civil

cases.

14

Recommendation #5: One way for New Jersey to show its

commitment to seating more representative juries would be to

increase the size of civil juries to 12 members, rather than six. A

recent article by Federal judges argues for the superiority of the 12-

person jury, and the data in this study amply demonstrate why larger-

sized groups do a better job than smaller-sized groups of achieving

stable levels of representativeness and avoiding the wholesale

elimination of minority-group members.

Section V. Peremptory Challenges

Given the sizeable allotment of peremptory challenges New Jersey grants to

parties, particularly to the defense in criminal cases, this study revealed a series of

surprising, counterintuitive findings. First attorneys’ use of peremptory strikes on

minority group members played only a case-specific and generally attenuated role

in explaining patterns of underrepresentation on juries. Attorneys rarely exercised

more than one strike against an African American venireperson in a trial, and

peremptory challenges failed to explain outcomes in cases in which African

Americans were wholly eliminated from juries. Patterns of peremptory challenges

did not explain many of the examples of concerning levels of underrepresentation

presented in Section IV. Latinos were somewhat more likely than African

Americans to be struck by a peremptory challenge, but again, peremptory use did

not have much purchase in explaining why some juries failed to have any Latinos

(or Asians), nor did they regularly account for the instances of underrepresentation

described in Section IV.

15

Analyses that examined whether peremptories disproportionately affect non-

Hispanic Whites found that criminal defense attorneys were about one-quarter as

likely to use a strike on an African American prospective juror as on a White

prospective juror. However, that was the sole statistically significant effect. In

general, the data show that attorneys do not use a substantial number of challenges,

and only rarely use all or close to their full complement of strikes. As an empirical

matter, the data indicate that prosecutors in all but the rarest of cases needed 8

strikes, and likewise the vast majority of criminal defendants used but 10; on the

civil side, six strikes in total would have covered all but a handful of parties in

these trials.

Nonetheless, as New Jersey works on deciding the “right” number of

peremptory challenges to permit, I do not recommend relying solely on the

empirical account of how attorneys used their strikes. Understanding peremptory

use in these trials requires a careful look at another source of attrition from jury

service – judges’ generous use of challenges for cause – because, very likely,

peremptory use is responsive to judges’ behavior. Results in Section VII, which

review the remarkably large venire sizes in New Jersey trials – particularly in

criminal cases – provide a stronger empirical justification for reducing peremptory

challenges compared to concerns about the peremptory’s effect on levels of

representativeness.

16

Section VI. Challenges for Cause

In New Jersey criminal trials, challenges for cause are the primary means of

attrition from a venire, removing over half of all venirepersons. In civil cases,

cause challenges are second, just behind the “not used” category, in ending service

for people, and they account for just under 40% of the exits from jury service. No

evidence in this study indicates that attorneys need a large number of peremptory

strikes because judges are reluctant to grant reasonable cause challenges. Were this

claim plausible, one would expect to see a “negative association” between

peremptory strikes and cause challenges; that is, attorneys would use more

peremptory challenges as judges excused smaller proportions of the venire through

cause challenges. Far from using more strikes to offset judges’ conservative use of

challenges, there was in fact a positive association between cause challenges and

peremptory challenges. Thus, both tend to increase or decrease in tandem across

cases, likely in response to venire characteristics. The significance of this finding

for understanding peremptory use in New Jersey cannot be overstated. The liberal

use of challenges likely explains why attorneys do not often use their full

complement of challenges.

Although judges’ challenges for cause account for substantial proportions of

jury selection outcomes, these challenges are not statistically related to juror race

and ethnicity. By and large, judges grant cause challenges to members of different

17

racial/ethnic groups at comparable rates. Finally, I examined whether active use of

cause challenges affects jury diversity. In criminal cases, there was no statistically

significant relationship between cause challenges and the proportion of non-

Hispanic Whites on the final jury – a rough but plausible measure of jury diversity.

In civil cases, there was a surprising, negative statistical association: the higher the

proportion of people dismissed through challenges for cause, the lower the

proportion of Whites on resulting juries. In all likelihood, other factors not

measured in this study explain the effect, but it is further evidence that judges can

liberally grant challenges for cause without concern that doing so undermines jury

diversity.

Recommendation #6. Judges in New Jersey courts should continue

their current practices in granting challenges for cause when they

deem them to be appropriate. The rate at which New Jersey judges

grant challenges for cause, while high, arguably allows attorneys to be

conservative in their use of peremptory challenges, and their use does

nothing to undermine jury diversity.

Section VII. The Size of Jury Venires

The current study reveals a number of positive aspects of the jury selection

process in New Jersey: Most basically, New Jersey took the initiative to study its

practices by fielding the study that produced these data. Second, although New

Jersey resembles most jurisdictions by substantially underrepresenting minorities

(particularly African Americans) in the pools of people who appear for jury

18

service, and although there were pockets of concerning levels of minority

underrepresentation on juries, in the main I find that New Jersey does an admirable

job producing juries that mirror their venires. Third, despite statutes that grant

attorneys the opportunity to use a large number of peremptory challenges

(especially in criminal cases), the data showed a fairly constrained use of these

strikes, and only limited evidence that attorneys’ peremptories negatively affect the

representativeness of the final juries. Finally, New Jersey’s judges appear to

liberally grant challenges for cause, and no evidence suggests that these strikes

have a disparate impact on any racial/ethnic group. Judges’ approach to these

strikes may explain attorneys’ constrained use of challenges, and additionally,

judges’ challenges do nothing to undermine jury diversity.

Nonetheless, one aspect of jury selection practice in New Jersey is less

commendable. In criminal cases, the second-most common way that people ended

their voir dire experience was by being “not used,” and this was the most common

way that people in civil cases ended their appearance at a voir dire. In a few

counties, the not-used group accounted for more than 50% of outcomes in the

venires, and in one of these areas, “not used” made up more than 60% of venires.

Venires in criminal cases averaged 144 people, which is substantially higher than

reported sizes from other jurisdictions. Although the “right” size for a venire will

vary by the needs of cases and an understanding of a jurisdiction, in these data the

19

general picture is that venires pull together more prospective jurors than is

necessary to intelligently exercise challenges for cause, exercise peremptory

challenges, and seat a jury, particularly in criminal cases. The excess number of

jurors who do not fall into any of these categories – that is, who wind up being not

used – does nothing to enhance the diversity of final juries and risks making

citizens feel their time is wasted, which undermines the legitimacy of courts.

The size of venires likely contributes to the high proportions of challenges

for cause reviewed in Section VI. Indeed, in criminal cases, the size of the venire

was strongly and positively associated with the rate of cause challenges – the

bigger the venire, the higher the rate of people excused for cause. In civil cases, in

which the range of venire sizes is more attenuated, the association was less strong

and only marginally significant but was still positive in direction. Thus, the aim

should be to reduce the size of the venire without reducing judges’ comfort with

dismissing people the judge believes cannot be fair and impartial in a case.

Of note, in both civil and criminal cases, the absolute size of a venire was

uncorrelated with the diversity of the final panel. Therefore, large venires, by

themselves, do not accomplish diversity goals. Instead, the best correlate of a

diverse jury is a diverse venire, yet another reason why New Jersey should do all it

can to improve the representativeness of those who appear at court, particularly

African Americans, but certainly all groups (see Section III and Recommendation

20

#3). But for the sake of the resources of both courts and of people who already

appear for service, New Jersey should critically examine why so many people are

called to a voir dire only to wind up “not used.” If planning for attorneys’ using

large numbers of peremptory challenges contributed to the large panel sizes in

these trials, then this is a strong, empirically supported reason to reduce the number

of challenges allotted, particularly in criminal cases. Alternatively, or in addition, if

judges’ willingness to support challenges for cause depends on having large panels

to avoid “running out” of prospective jurors, then New Jersey should investigate

the minimum amount of surplus necessary to permit judges to be comfortable

granting cause challenges they deem appropriate. The latter would achieve the goal

of lowering venire sizes while preserving a practice that appears to constrain

peremptory challenge use and does not harm jury diversity. Thus, my final

recommendation to New Jersey is to focus scrutiny on the size of venires.

Recommendation #7: New Jersey should determine ways to reduce

the number of people who are called to voir dire only to be “not

used.” Possible mechanisms include reducing the number of

peremptory challenges, particularly in criminal trials, or convening

judges to consider by how much venire panels might be reduced while

still allowing judges to be comfortable during cause challenge

determinations.

Further details on all the elements of this summary follow. Throughout,

findings I regard as particularly notable appear in bolded text.

21

SECTION I. A STUDY TO OBSERVE RACE, ETHNICITY, AND GENDER

IN NEW JERSEY COURTS

Prior reports have detailed the process that the New Jersey Courts Office

developed to study the race, Hispanic/Latino ethnicity, and gender of people who

appear for jury service.2 In brief, during a several-week period in September and

October of 2018, fifteen counties disseminated one-page questionnaires to all

persons appearing for jury service. Through the questionnaire, people could

voluntarily indicate their race,3 Hispanic/Latino ethnicity,4 and gender (male or

female). Each questionnaire was affixed with a person-specific bar code number,

and this matched the person’s number on his/her juror badge. This permitted

2 See “Memorandum re: Study Design and Implementation Plan Proposal,” from Jessica

Lewis Kelly and Lisa R. Burke, April 23, 2018; and “Final Report on the Combined Jury

Studies on Jury Representativeness and the Impact of Peremptory Challenges on the

Racial and Ethnic Composition of Petit Juries (September 2018 – October 2018),” by

Brian J. McLaughlin and Lisa R. Burke.

3 Categories presented to respondents (with descriptions of areas of origin) were: White

or Caucasian, African-American or Black, American Indian or Alaskan Native, Asian,

Native Hawaiian or Pacific Islander, Other, and Multi-Racial. A respondent was

permitted to check more than one category (“check all applicable categories”), and a note

at the top of the page indicated that answers to all questions were voluntary (“Your

voluntary participation is requested”).

4 Categories: “Hispanic or Latin American” (with description of types of origins) and

“Not Hispanic or Latin American.” The questionnaire indicated that these were based on

“U.S. Census Bureau definitions.”

22

tracking and matching questionnaires to information on outcomes stored in the

Jury Management System (JMS) software.

Thus, for those who filled out a survey and had it collected at the

courthouse, New Jersey obtained de-identifiable information on these persons’

race, Latino ethnicity,5 and gender, as well as their outcome at the close of their

appearance. These outcomes included whether they were part of a pool only vs.

going to voir dire and, among the latter group, whether they were selected as juror,

excused for cause, peremptory challenged by the plaintiff (civil) or prosecutor

(criminal), peremptory challenged by the defendant (civil or criminal), or not

reached during voir dire (coded in the data as “not used”). For those not filling out

a questionnaire, we know only their outcome through the JMS system.

As an expert in jury representativeness issues, the New Jersey Supreme

Court appointed me to analyze the de-identified data to examine, at minimum, the

5 People who self-identify in the “Hispanic or Latin American” category use a wide

variety of terms to describe themselves. Common are: Hispanic, Latino, Latinx, Chicano,

Mestizo, or any country-specific identifications (e.g., Cuban-American, Mexican

American, etc.). Many terms are regional (e.g., Chicano tends to be more common in

West-Coast/Southwestern areas), or may be used to match governmental usage (e.g., the

Census Bureau uses “Hispanic Origins” or “Hispanic or Latino”). Recognizing that

people vary a great deal in what term they prefer, this report generally uses the term

“Latino” or “Latino/a” (which is more inclusive of gender). Although a newer, related

term, “Latinx,” has emerged, currently few people self-report using this term (see

https://www.theatlantic.com/ideas/archive/2019/12/why-latinx-cant-catch-on/603943/,

citing poll results indicating that just 2% of Hispanic/Latino respondents embraced the

term “Latinx”). In this report, I use “Black” and “African American” interchangeably,

and “Asian” and “Asian American” interchangeably.

23

following questions: (1) how racially and ethnically diverse are the jury pools in

the participating counties in this study? (2) what patterns of attrition are observed

for those people who undergo voir dire for a civil or criminal case? and (3) do

patterns of attrition exist that correlate with a person’s race, ethnicity, or gender?

The Court indicated a particular interest in whether the existing system of

peremptory challenges, and their use in trials, substantially altered the likelihood

that members of various racial or ethnic groups were seated on juries.

A. Cleaning data. After I, and my colleague, Marc A. Musick, Ph.D., who

has extensive experience working with large datasets, signed appropriate

paperwork to serve the State of New Jersey and to maintain the confidentiality of

the data, we received access to a de-identified dataset. All data we used identified

each observation by only their bar code, and we had no other identifying

information on people in the dataset beyond their responses to the questionnaires

and their jury service outcome. We were first provided an FTP link to a Microsoft

Access file prepared, as we understand it, by information technology (IT)

personnel who oversaw the process of getting all the bar-coded forms into the

dataset. Upon downloading the data, checking for duplicate records, and arranging

the entries into unique county/trial numbers, we identified problem cases and

reported these to Brian McLaughlin, who together with Lisa Burke, served as our

primary contact on study issues. After they researched the cases, we learned of

24

clear discrepancies between the data downloaded through the Access file and

information in the larger JMS records the counties kept (e.g., a study-eligible trial

would be listed in the Access file as having 11 jurors, despite the JMS records

indicating there were 14 jurors).

To remedy this, we began again with a new set of data files: all the raw

survey-data files from each county and all the JMS files from each county. Dr.

Musick worked intensively with each file to merge survey and JMS records, to

clean the data, and then to combine all county-level files into one master dataset.

The data-cleaning process involved identifying which observations were out of the

study’s date range, which JMS records did and did not have surveys to match,

which observations reflected the same person but two different outcomes (e.g., a

person went to voir dire for two different trials during his or her time at the

courthouse),6 and which observations were wholly redundant and needed to be

deleted (e.g., a single person filled out a survey on more than one occasion but did

not experience two different outcomes). Purged of true redundancies (but not

instances of the same person having multiple outcomes) and entries that fell out of

the date range of the study (e.g., trials from December of 2020), the cleaned dataset

included 18,668 total observations, representing 149 unique trial numbers together

6 This is evident in the data when the same bar-code number is associated with two

different trial numbers.

25

with 15 sets of people who had no trial number because they served only in the

“pool” and never underwent voir dire.

Because in some counties the same person could encounter more than one

trial during an appearance at court, there were more observations than individual

people in the dataset. This initial cleaned dataset reflected 15,529 unique

individuals: 81% of whom, or 12,525, had just one trial- or pool-only experience;

2,870, or 18% had two different trial encounters; and 134, or 1%, of people had

three different trial experiences. Other issues further reduced the size of the dataset

(described in Section II and in detail in Appendix A), but the cleaned dataset of

18,668 observations and 15,529 people represents the baseline dataset of persons

who, according to JMS records, appeared at the courthouses studied during the

dates of the study. Within that database are the subset of people who filled out

questionnaires that the courthouses collected.

B. Conclusion and Recommendation #1. As the above description reveals,

the first – and exceedingly clear – conclusion is that this study was incredibly

labor-intensive. Consider all the demands this project placed on various types of

personnel: the work that went into designing a creative and confidential way to

collect the data; pre-testing the procedures; the need to enlist county-level jury

offices and jury clerks to run the study when they were likely to be busy with a

number of other tasks; the time it took for IT personnel to transform raw survey

26

data into usable files; initial in-house examinations of the data and report-writing;

and the roughly 25 hours two Ph.D.-level social scientists spent solely on

transforming and cleaning the data to make it trustworthy for analysis,7 apart from

additional statistical analysis and writing.

This massive effort demonstrates New Jersey’s commitment to meeting

both its constitutional obligations and its own policy values concerning jury

representativeness,8 thus making New Jersey a leader among states in this

arena. However, New Jersey could meet its obligations and policy values in a

far more efficient manner and should develop a system to routinely measure,

at minimum, the race, ethnicity, and gender of all persons appearing for

service. These categories represent the types of “cognizable groups” recognized in

jurisprudence on representativeness;9 however, questionnaires could be modified

as needed to also measure any other characteristic the Judiciary believes

contributes to healthy diversity on juries.

7 The hours attending to problems with the data consumed the majority of the 36 hours

initially allotted for expert examination of the data. Although an additional 20 hours were

approved, the work necessary to complete this report went well beyond the total allotted,

amounting to approximately 100 additional hours.

8 See, e.g., “Memorandum re: Study Design and Implementation Plan Proposal,” from

Jessica Lewis Kelly and Lisa R. Burke, April 23, 2018, at page 1.

9 See, e.g., Duren v. Missouri, 439 U.S. 357 (1979); Berghuis v. Smith, 559 U.S. 314

(2010).

27

Courts should collect information on cognizable groups in a fashion that

permits regular in-house analysis to quickly and reliably check for problems in

representativeness. This would avoid fielding a complex study that disrupts normal

jury operations, and if the system were designed correctly, New Jersey would not

have to rely on substantial and costly outside assistance in order to ask basic

questions about the racial identity of people who appear at court for jury service.

New Jersey is not alone among states in failing to collect jurors’ demographic

data;10 however, this failing should not and need not continue. New Jersey should

develop a process that is appropriate for its own summoning system,11 identifying

ways to make it easy for prospective jurors to report their background

characteristics to courts and easy for New Jersey to store and review these data.

Further, consistent with Supreme Court rulings that permit parties to have

access to data in order to pursue or defend claims in challenges to jury

representation, 12 data on cognizable groups should be stored in a way that allows

parties to access either de-identified data or reports based on such data for purposes

10 See, e.g., Nina Chernoff, No Records, No Right: Discovery and the Fair Cross-Section

Guarantee, 101 Iowa Law Rev. 1719 (2016).

11 Federal courts, for example, use a two-step jury selection process, and, at the first

stage, send qualification questionnaires to people on its jury wheels that include questions

about race and Latino ethnicity on these forms.

12 Test v. U.S., 420 U.S. 28, 30 (1975); see generally, Chernoff, supra, note 10.

28

of considering or pursuing a claim about the representativeness of jury pools.

Given the complexity of the above study, it is unfair to burden parties with trying

to collect racial or ethnic data themselves.

Speaking as a scholar in this area, ideally de-identified data could also be

made publicly available for those making reasonable requests in the public interest.

This transparency ensures public trust in the courts and promotes scholarly

understanding of the effects of court practices on jury representation levels. As this

report will reveal, there is much that New Jersey does well in securing racially

representative juries. A reliable and usable system for collecting and

analyzing, at minimum, the race, ethnicity, and gender of all who appear for

jury service constitutes the best way to ensure that the state continues its

successes, and it would allow the state to regularly examine practices that are

not working as well. A system that easily permits New Jersey to regularly monitor

and improve its practices is superior to a one-time, challenging study in conveying

New Jersey’s commitment to securing racially representative juries. By

disseminating such data (again, de-identified), an in-house system would also

permit the state to regularly publicly communicate that commitment and to

disseminate its findings for the benefit of the public in New Jersey and other areas.

Recommendation #1: New Jersey courts should develop a means of

asking all people who appear at the courthouse to identify how they

categorize themselves in terms of their race, Latino/a ethnicity, and

gender. This should be incorporated into any existing questionnaires

29

prospective jurors may fill out, or New Jersey should explore a new

means of getting this information from all individuals. Such data

should be used for in-house analyses of the effect of court practices on

jury representativeness, and de-identified data (or detailed reports

based on the data) should be available to parties pursing claims about

levels of jury representativeness. New Jersey Courts should also

consider making data available to scholars or members of the public

seeking to understand the effect of court practices on representation.

30

SECTION II. RESPONSE RATES, DETERMINING USABLE DATA, AND

MEASUREMENT ISSUES

In the dataset, trial numbers indicated a unique trial event.13 As we cleaned

the data, we noted that several trials commenced voir dire before the survey period

began; in addition, some trials continued after the survey period ended.

Information on these trials was incomplete (e.g., the file might list just 9 jurors for

a criminal case, or 4 jurors for a civil case). After confirming with the

Administrative Office that no additional information from these 31 trials would be

forthcoming, we omitted all of these trials as incomplete.

A. Response rates. Among usable trials, other forms of missing data were

present. First, counties varied in the rate at which they were able to produce survey

data from prospective jurors. I calculated the response rate as the number of

observations with survey responses on the race/ethnicity/or gender questions

(including responses in which people refused to answer a question) divided by the

total number of observations in the JMS data files (i.e., total persons recorded as

present at the courthouse for the same trials or time periods). Table II.1 presents

the response rates for each county.

13 Individuals who appeared at court but were only part of pool and never experienced

voir dire also had entries in the trial numbers field, but these were easily distinguished

from actual trial numbers.

31

The overall response rate to the study was 76.8%, indicating that on average

about three-quarters of participants filled out and returned questionnaires. Most

counties produced high response rates, or, at minimum, provided a sample that

constituted a clear majority of all observations from the county: Bergen,

Cumberland, Gloucester, Morris, Passaic, Somerset, and Union all had response

rates to the survey above 90%; Camden, Mercer, Middlesex, and Ocean counties

also had rates above 80%. Burlington, Essex and Monmouth had response rates of

two-thirds to 71%. However, for reasons that are not well understood, Hudson

County was an outlier, with a response rate of just 33%, which was half the rate of

counties with the next-lowest response (see Table II.1). I therefore had to omit

Hudson because any analysis of the demographic composition of pools, venires,14

or juries would be subject to high levels of error given the levels of missing data.

(Without Hudson County included, the overall survey response rate is 84.7%).

Even within the remaining counties, the number of observations that had no

surveys varied. One civil trial from Burlington County had unusually high levels of

missing data: 83% of those in the jury pool had no survey, and 86% of seated

14 Terminology regarding jury selection varies a good deal. Chernoff, supra, note 10, for

example, calls “venires” the group of people who appear at a courthouse at all in a given

period; in this report, I refer to that group as a “pool.” I use “venire” to refer to the people

who go to a particular courtroom for voir dire. By contrast, Professor Chernoff terms this

group a “panel,” which I use interchangeably with “venire.”

32

Table II.1: Response Rates, and Responses to Race/Ethnicity Questions, Across

Counties of Study.

Total

Total

with

Response

Missing on

Race

Missing on

Ethnicity

County Records Surveys Rate N Percent* N Percent*

Bergen 1,413 1,319 93.4 48 3.4/ 3.6 146 10.3/ 11.1

Burlington 1,095 744 67.9 13 1.2/ 1.7 110 10.0/ 14.8

Camden 958 900 84.0 11 1.2/ 1.2 73 7.6/ 8.1

Cumberland 256 249 97.3 18 7.0/ 7.2 36 14.1/ 14.5

Essex 2,038 1,449 71.1 34 1.7/ 2.3 171 8.4/ 11.8

Gloucester 387 362 93.8 4 1.0/1.1 50 12.9/ 13.8

Hudson 2,382 796 33.4 36 1.5/ 4.5 82 3.4/ 10.3

Mercer 903 805 89.2 14 1.6/ 1.7 113 12.5/ 14.0

Middlesex 1,374 1,133 82.5 38 2.8/ 3.4 173 12.6/ 15.3

Monmouth 1,016 689 67.8 9 0.9/ 1.3 72 7.1/ 10.4

Morris 766 742 96.9 18 2.3/ 2.4 65 8.5/ 8.8

Ocean 676 577 85.4 8 1.2/ 1.4 68 10.1/ 11.8

Passaic 1,161 1,119 96.4 46 4.0/ 4.1 114 9.8/ 10.2

Somerset 387 376 97.2 6 1.6/ 1.6 19 4.9/ 5.1

Union 717 668 93.2 52 7.3/ 7.8 98 13.7/ 14.7

Total in Data 15,529 11,928 76.8+ 355 2.3/ 3.0 1,390 9.0/ 11.7

Totals after

Omissions+ 12,462 11,132 84.7 -- -- -- --

+See text for explanation of omitting Hudson County and trials with incomplete information.

*Value before the slash is the percent of total records, value following the slash is the percent of

those who filled out surveys. See Appendix A for information on how missing data were

managed.

jurors had no survey. It also indicated that no cause challenges and no peremptory

strikes occurred, rendering this case unusable in several types of analyses; I

therefore omitted it. This rate of missing data was an outlier case within

Burlington, but for Essex and Monmouth Counties – again for reasons that are not

33

clear – substantial levels of missing data appeared across most trials. In Essex,

30.3% of members of all the venires for trials had no survey (range: 23 – 41%), as

did 34% of all seated jurors (range: 13 - 67%); in Monmouth, comparable figures

were 34% (range: 19 – 46%) for venires and 39% (range: 14 - 71%) for juries.

Because these trials typically provided data on peremptory and cause challenge

usage, I did not eliminate them. Appendix A describes why I believe that

race/ethnicity values are still plausible for these counties, despite the presence of

missing data.

Following the elimination of trials with incomplete data, trials from Hudson,

and the outlier trial from Burlington, there remained 95 trial numbers across 14

counties. Of these, 26 (27%) were criminal, and 69 (73%) were civil. Thus, during

the study period, counties produced far more civil than criminal trials. Nonetheless,

owing to far larger jury pools in criminal trials, the group of criminal trials

generated slightly more individual-level observations in the data (3,753) than did

the civil cases (3,654).

B. Final dataset. This set of 7,407 observations involving a civil or criminal

voir dire/trial experience, together with the 5,055 people across the 14 counties

who were “pool only” (no voir dire experience), constitute the dataset that

produced the findings in this report. These 12,462 observations represent 12,263

34

unique people, of whom 10,358 also had questionnaire data about race, Latino

ethnicity, and gender.

C. Recommendations #2a., #2b, and #2c regarding future measurement

efforts. Appendix A, “Handling Missing Questionnaire Responses and Missing

Questionnaires,” describes additional choices I made when analyzing the data,

given that some observations lacked survey data, and as Table II.1 describes, some

people skipped questions about Latino identity or about race. In addition, people

were invited to select multiple categories in the race question, and some did.

Should New Jersey replicate this study or, as Recommendation #1 advocates,

should it design a system to routinely measure race and ethnicity, Appendix A

offers a technical primer for this area of study. It provides the bases for a series of

recommendations regarding optimal questionnaire design and measurement of race

and ethnicity, particularly in light of the law’s demand that representation concern

a “cognizable group.” For simplicity, I restate them here:

Recommendation #2a: Because of the policy and legal importance of

understanding the demographic patterns in a jury selection system,

people should not be specifically invited to consider questions about

race, ethnicity, or gender to be voluntary. This will lower the

likelihood that people fail to turn in a survey at all or that they skip

questions (e.g., by indicating that the question is optional). New

Jersey may take a position that such questions should be voluntary. If

so, court personnel can treat the survey that way by not admonishing

people who fail to turn one in, not returning questionnaires to people

who have skipped questions, and by programming any online

questionnaires so that people can skip a question. But neither the text

of the questionnaire itself nor the race/ethnicity questions should

35

invite people to fail to answer. Instead, people should be informed

about the reasons why the questions are necessary to ask, with a brief

explanation accompanying any questions about race, ethnicity or

gender.

Recommendation #2b: Because the concept of a “cognizable group”

is one requirement for proving that underrepresentation has occurred

in the jury summoning, qualification, or selection process, questions

about race should not invite people to select multiple categories.

Instead, respondents should be asked to select the category that “best

describes” their race. Options appearing on the form can include a

“Multiracial” category so that people who feel that this designation

best describes them can select it. However, a “best-describes” design

would increase the likelihood of identifying those people who may be

multiracial but who tend to identify with and experience their race

primarily through one aspect of their background more than another; it

also reduces instances in which people are part of a group that is too

small in size to reliably analyze.

Recommendation #2c: To minimize the tendency to skip a question

about Latino ethnicity whenever it follows rather than precedes a

question about race, respondents should be asked about whether or not

they are Latino/a before they are asked to identify the racial group that

best describes them.

36

SECTION III. COUNTY RACIAL AND ETHNIC DEMOGRAPHIC

PROFILES.

A. Absolute and comparative disparities across counties. Although this

report devotes most of its attention to patterns of jury participation and attrition

once people appear at court, a key question ahead of that is how well groups

appearing at court represent the areas from which they are drawn. This section

reports on the racial and ethnic composition of the people in the dataset compared

to each county's actual demographic composition. The American Community

Survey (ACS) constitutes the best source of information on the jury-eligible adult

population for each of the 14 areas that produced data for the study.15

15 Beginning in the mid-2000s, the ACS sampled areas in the years between each

decennial census. The ACS is advantageous because it keeps current with changes in the

population between decennial censuses, and it also includes a question on citizenship.

This provides a strong estimate of the jury-eligible population in an area (i. e., citizens

who are 18 and over). I used the five-year estimates, which are the most precise and

widely available values. They are precise because they combine samples across years,

which reduces the size of sampling error for any one year. They are the most widely

available because, for smaller areas or areas with fewer minorities, the five-year estimates

are the only information reported (i.e., to have enough people to develop plausible

estimates of small sub-groups). Because it merges across previous years, results represent

a conservative estimate for the population profile of groups specifically in 2018 (e.g., the

size of the group may have increased in a given area across the five-year period). These

estimates are available from the Census Bureau website (Table B05003; see

https://www.census.gov/acs/www/data/data-tables-and-tools/data-profiles/).

37

Table III.1: Racial/Ethnic Profile of Counties in Study, American

Community Survey Data, 2014 – 2018, Table B05003 (Adult

Native-Born and Naturalized Citizens)

Percentage who are:

County

Total in

Area Black Latino/a Asian Whitea

Bergen 635,116 6.3 16.1 12.4 64.5

Burlington 336,960 16.4 6.3 3.6 71.7

Camden 369,232 19.2 11.7 4.6 63.5

Cumberland 107,262 20.9 20.4 1.3 55.6

Essex 517,433 41.3 17.3 4.6 36.0

Gloucester 221,039 10.2 4.4 2.4 81.9

Mercer 250,516 20.8 10.3 8.1 59.8

Middlesex 537,502 11.0 15.9 17.3 54.8

Monmouth 457,917 6.8 6.6 4.6 81.1

Morris 353,867 3.5 9.3 7.7 78.7

Ocean 433,901 3.1 5.8 1.5 88.9

Passaic 324,989 12.6 31.4 4.8 50.5

Somerset 225,865 10 10.2 12.8 66.3

Union 350,104 23.6 21.6 4.6 48.8

Total in Data 5,121,703 795,176 687,568 366,692 3,276,797

Overall Proportion 15.5 13.4 7.2 64.0

Notes: aNumbers are based on the “White/Non-Hispanic” category in the data.

The areas with the largest racial/ethnic minority populations are Essex,

Union, Cumberland, Mercer, Camden, and Burlington Counties, all of which have

African American populations that exceed 15% of the community. For Latinos,

Passaic, Union, Cumberland, Essex, Bergen, and Middlesex Counties exceed 15%.

For Asians, Middlesex, Somerset, and Bergen have populations that exceed 10% of

the area. Of all counties, only Union and Essex are not majority-White/non-

38

Hispanic. Using ACS data, Table III.1 shows how the counties break down along

racial and ethnic lines.16

Table III.2 on the next page reports the demographic profile of those persons

who provided questionnaire responses in the 14 counties in this study. Comparing

overall percentages listed in the bottom row of Table III.1 to those at the bottom of

III.2, African Americans are underrepresented in the aggregate of New Jersey jury

pools; they are 15.5% of these areas but just 12.1% of jury pools. This “absolute

disparity” of 3.4% (15.5 – 12.1), divided by the size of the population in the

community (15.5%), yields a “comparative disparity” of nearly 22%. In other

words, about one in five members of the African American population of New

Jersey is “missing” from jury pools. Latinos demonstrate a smaller absolute

disparity of 1% (just a 7% comparative disparity). Asians and Whites (non-

Hispanic) are overrepresented in the aggregate.

16 In addition to providing information on racial and ethnic profiles of each area, Table

III.A indicates that the 14 counties included in the study represent an excellent sample of

the entire state. According to the ACS, there are 6,117,615 citizens who are 18 and older

in New Jersey. The 14 counties studied have populations that together amount to over

five million people and constitute 84% of all adult citizens in New Jersey (that coverage

increases to 90% had we been able to use Hudson County’s data).

39

Table III.2: Racial/Ethnic Profile, Overall and by County, Based

on Respondents to the Survey

Percentage who are:

County

Total in

Data Black Latino/a Asian Whitea

Bergen 1,224 4.5 16.1 13.5 63.9

Burlington 744 15.5 4.4 3.9 73.5

Camden 788 12.8 9 3.5 73.6

Cumberland 249 14.9 17.8 1.6 62.3

Essex 1,391 32.9 15.9 7.5 42.3

Gloucester 271 6.3 2.2 2.6 87.1

Mercer 748 12.2 7.9 11.8 66.0

Middlesex 937 9.4 14.3 20 55.4

Monmouth 689 2.9 6.5 7.4 80.1

Morris 742 3.1 9.4 9.8 76.0

Ocean 543 2.0 7.9 1.0 87.7

Passaic 1,064 9.6 22.3 7.7 58.9

Somerset 376 4.8 9.0 17.0 68.6

Union 592 19.3 16.9 7.8 49.7

Total in Data 10,358 1,249 1,308 933 6,673

Overall Percentage 12.1 12.6 9.0 64.4

Notes: a In order to match Census bureau figures, numbers are based on people

in the data who said they were “White” but not “Hispanic.”

See Appendix A for how analyses managed instances in which people did

not answer the race or ethnicity question.

These aggregate-level disparities are not constant across counties. Table III.3

next summarizes the comparison of Tables III.1 and III.2 and reports both the

absolute and comparative disparities for each county and for each racial/ethnic

group.

40

Table III.3: Absolute and Comparative Disparities: All People Appearing for

Service Who Filled Out Surveys.

Absolute Disparities Comparative Disparities

County Black Latino/a Asian White Black Latino/a Asian White

Bergen (1.8) 0.0 1.1 0.6 (28.6) 0.0 8.9 0.9

Burlington 0.9 (1.9) 0.3 1.8 5.5 (30.2) 8.3 2.5

Camden (6.4) 2.7 1.1 10.1 (33.3) 23.1 23.9 15.9

Cumberland (6.0) 2.6 0.3 6.7 (28.7) 12.7 23.1 12.1

Essex 8.4 1.4 2.9 6.3 20.3 8.1 63.0 17.5

Gloucester (3.9) (2.2) 0.2 5.2 (38.2) (50.0) 8.3 6.3

Mercer (8.6) 2.4 3.7 6.2 (41.3) 23.3 45.7 10.4

Middlesex 1.6 1.6 2.7 0.6 14.5 10.1 15.6 1.1

Monmouth (3.9) 0.1 2.8 1.0 (57.4) 1.5 60.9 1.2

Morris 0.4 0.1 2.1 2.7 11.4 1.1 27.3 3.4

Ocean (1.1) 2.1 (0.5) 1.2 (35.5) 36.2 (33.3) 1.3

Passaic (3.0) (9.1) 2.9 8.4 (23.8) (29.0) 60.4 16.6

Somerset (5.2) 1.2 4.2 2.3 (52.0) 11.8 32.8 3.5

Union 4.3 4.7 3.2 0.9 18.2 21.8 69.6 1.8

Notes: Values in parenthesis reflect instances in which a group is underrepresented in jury pools,

compared to that group’s proportion in the community, by at least a 25% comparative disparity

(see Table III.1 for data on each county’s population profile and Table III.2 for jury pool results).

Values in italics indicate overrepresentation.

The first clear pattern in Table III.3 is that African Americans are

underrepresented in the jury pools to some degree in each one of the 14

counties studied. For some counties the level of underrepresentation of African

Americans is quite low, particularly Burlington, Middlesex, and Morris, all of

which have absolute disparities of less than 2% and comparative disparities below

15%. How might one decide whether a county has “a lot” or “a little”

41

underrepresentation? Some scholars have argued that a 15% comparative disparity

should signal “not fair and reasonable” underrepresentation, a component of the

requirements of proving underrepresentation in Duren.17 Burlington, Middlesex,

and Morris are the only counties that fall below this standard.

However, this particular standard is not only quite liberal,18 but also

complicated given that the standard’s advocates also suggest that communities

with smaller proportions of minority group members – that is, groups that make up

less than 10% of a community – should be held to a less strict standard, with

comparative disparities of 25% or more regarded as “not fair and reasonable.”19

Because the population prevalence of minority populations in this study varies so

much across the fourteen counties, I opted for a single standard to signal

concerning levels of underrepresentation: at least a 25% comparative disparity

between the population and the jury pools in that county. In Table III.3 I have

17 Duren v. Missouri, 439 U.S. 357 (1979). For scholars advocating a 15% comparative

disparity standard, see, e.g., Kairys, David, Joseph B. Kadane, and John P. Lehoczky,

Jury Representativeness: A Mandate for Multiple Source Lists, 65 California Law Rev.

776 (1977); “Brief for Social Scientists, Statisticians, and Law Professors,” Jeffrey

Fagan, et al., as Amici Curiae Supporting Respondent (2009), Filed in Berghuis v. Smith,

559 U.S. 314 (2010).

18 See Mary R. Rose, Raul S. Casarez, and Carmen M. Gutierrez, Jury Representation in

the Modern Era: Evidence from Federal Courts, 15 JOURNAL OF EMPIRICAL LEGAL

STUDIES 378 (2018) (finding that nearly all instances of underrepresentation in federal

districts had at least a 15% comparative disparity).

19 Kairys, et al., supra, note 17.

42

placed parentheses around those instances in which levels of underrepresentation

are associated with at least a 25% comparative disparity.20 (I will later use this

same standard for other analyses, such as the difference between venires and juries;

see Section IV.)

In addition to the three counties with low absolute disparities, Union County

is the only other area that has a comparative disparity below 25%, although just

barely (see right side of Table III.3). Gloucester, Mercer, Monmouth, and Somerset

have the largest comparative disparities, and for most of these areas, this cannot be

20 Scholars favor the comparative disparity because, of all indicators, it is not only simple

to calculate but also considers underrepresentation in light of a group’s proportion in the

population. Values indicate the proportion of that group’s population that are “missing”

from jury pools due an absolute disparity (see, e.g., “Brief for Social Scientists,

Statisticians, and Law Professors,” supra, note 17). By contrast, the absolute disparity

conveys less information. For example, a 5% absolute disparity could mean that the

entirety of a small (5%) population has been eliminated from jury pools, or it could

indicate that a group that makes up 55% of the population is only 50% of an area’s jury

pool, and therefore, that group’s presence in jury panels is only minimally reduced.

Nonetheless, the Supreme Court indicated in Berghuis that jurisdictions should examine

all tests of underrepresentation when assessing underrepresentation. Thus a 25%

comparative disparity may not be concerning when the absolute disparity is very low

(e.g., Asians are underrepresented in Ocean County by less than 1%, but this generates a

comparative disparity of 33%; see Table III.3). A different test of underrepresentation

examines the statistical significance of these disparities – known as the “SDT” test (see

Berghuis v. Smith, 559 U.S. 314, 2010). I do not apply that test in this section because it

is highly sensitive to sample size, and most jurisdictions have surveys from hundreds of

people (see Table III.2). This tends to make even small differences “significant” (see,

e.g., Rose et al., supra, note 18). I use statistical significance tests for venire outcomes,

but as I describe in Appendix B, statistical tests of venire outcomes tend to be

conservative because both selection to a jury and other outcomes, such being subject to a

peremptory challenge, happen to only small proportions of a venire (see Appendix B for

more detail).

43

attributed to the fact that they have particularly small African American

populations in their counties, which can lead to exaggerated comparative disparity

figures: all but Monmouth have African American populations of at least 10% (see

Table III.1). In sum, although a few counties had low levels of

underrepresentation of African Americans among those appearing at

courthouses for jury service, rates were at concerning levels in fully 10 of the

14 areas studied.

The underrepresentation of Latinos had more county-specific patterns.

Just three areas had comparative disparities above 25%, and only one of these

(Passaic) also had a very sizeable absolute disparity (9.1). The other high

comparative disparities in Gloucester and Burlington owe to small absolute

disparities in areas with a low prevalence of Latinos. Two areas (Ocean and

Morris) had an overrepresentation of Latinos in their jury pools (indicated in Table

III.3 through italic text). As was the case in analyses of African American

representation levels, Union County’s underrepresentation of Latinos was just shy

of a 25% threshold (a 20% comparative disparity).

44

There is little evidence that Asians are systematically underrepresented

in jury pools in these data; only Camden and Ocean exhibited

underrepresentation, and both have Asian populations of less than 5%.21

B. Conclusions and Recommendation #3. The data make clear that

underrepresentation of African Americans is commonplace, with just three to

four counties demonstrating rates of African Americans in their pools that were

close to proportions in their communities. Unlike patterns for other racial/ethnic

groups, no county saw even slight overrepresentation of African Americans among

those appearing at court. By contrast, for Latinos, although a few specific areas can

do more to improve rates of this community’s appearances in its jury pools, the

general pattern of underrepresentation was not as consistent as it was for African

Americans.

Given that the data I analyzed concerned only those people who appeared at

the courthouse, I have no empirical tools to explain why African Americans are

frequently and substantially underrepresented in jury pools across counties,

21 This report focuses on racial and ethnic diversity. I looked for indications that women

or men are underrepresented in New Jersey’s jury pools and found little evidence.

Women are 52% of the population across these counties, with the percent of females in

the population ranging from 49% (Cumberland) to 53% (Essex). By comparison the

prevalence of women in the jury pools ranged from 47% (Morris) to 55% (Monmouth),

with no comparative disparity coming even close to a 25% standard. According to

statistical tests similar to those described in Appendix B, gender was also unrelated to

venire outcomes.

45

although this result is consistent with other studies of representation.22 This form of

attrition is an exceedingly difficult issue for courts to address. It is, for example,

multiply determined, traceable to potential biases in source lists, undelivered or

lost mail, summons non-response, and the fact that sometimes people forget that

they must appear.23 Apart from investigating the adequacy of source lists and

developing a means for reminding people about their appearance, most pre-

courthouse issues can be far from courts’ power to control (e.g., people moving

without updating their address with the Postal Service, thereby creating

undeliverable mail that cannot be easily updated in the system).

Further, one source of attrition can be costly to address. Courts could, for

example, more actively enforce a summons and follow-up with those who do not

appear. Greater oversight of summons nonresponse would not only be resource-

intensive but would also unleash more law enforcement into communities that, in

the current era, may already feel over-policed. At the same time, if New Jersey’s

goal is to create juries that better represent their communities, then it must better

understand and address sources of attrition in who makes it to the courthouse for

22 See review in Mary R. Rose, & Jeffrey B. Abramson, Data, Race, and the Courts:

Some Lessons on Empiricism from Jury Representation Cases, 2011 MICH. ST. L. REV.

911 (2011).

23 Bowler, Shaun, Kevin Esterling, and Dallas Holmes, GOTJ: Get out the Juror, 36

POLITICAL BEHAVIOR 515 (2014).

46

jury service, which may include the need to reach out to people in some manner to

ensure respect for the courts’ orders to appear for service, either through greater

use of reminders about a service date,24 or some form of enforcement. Changes to

procedures at the courthouse will yield only so much success in improving the

representativeness of juries if the basic disparities between pools and the broader

community are not addressed.

Recommendation #3: New Jersey’s next intensive study of jury

representativeness should be aimed at understanding the source(s) of

why jury pools – the groups of people who appear at the courthouse

for service – consistently and substantially underrepresent African

Americans. New Jersey should consider how to improve summons

response in ways that are appropriate for a given community and that

actually generate yields in participation, but which minimize costs to

the court and to the underrepresented communities. New Jersey may

need to enhance its system for reminding people about their assigned

service date and explore reasonable methods of summons

enforcement.

24 Id.

47

SECTION IV. RACE, ETHNICITY AND SERVING ON A JURY IN NEW

JERSEY COURTS

This section presents results that bear on the key question that motivated

New Jersey’s study: When people undergo voir dire, what is the likelihood that

service on a petit jury depends upon their racial or ethnic identity? The sample for

these analyses includes the 7,407 observations associated with those who

underwent voir dire for at least one trial – called the “venire” to distinguish it from

the “pools” discussed in Section III. Of venire-members in this study, 6,435 (87%)

also had questionnaire data on race or Latino ethnicity.25

A. Jury composition. Table IV.1 presents the proportions of each

racial/ethnic group in the venires compared to that group's proportion on juries for

each county. This table combines information across both civil and criminal trials

because doing so provides a larger, and therefore more reliable, sample. (Tables

IV.2 and IV.3, reviewed below, present the same comparisons separately for

criminal versus civil trials.)

25 As Appendix A describes, those without questionnaires were not included in

calculations of the racial/ethnic composition of venires and juries because there is little

evidence that questionnaire response correlated with outcome types. However, I

nonetheless examined the data with and without the inclusion of those who had

questionnaire data, and patterns of underrepresentation (e.g., whether a group met a

threshold of a 25% comparative disparity) were the same.

48

Following the convention used in Section III, parentheses in Table IV.1

indicate instances in which a group was underrepresented on juries, compared to its

proportion on the venire, by a comparative disparity of as much as 25%.26 For

African Americans, three counties had sizeable differences between

representation levels on juries compared to venires: Burlington, Camden, and

Somerset. Subtracting the percentage on juries compared to venires, the absolute

disparities were -6.4%, -6.8%, and -4.9% for these three counties respectively (the

associated comparative disparities are -39.8%, -54%, and 100%).27 These levels of

underrepresentation are substantial (in absolute and comparative terms), and are

additionally concerning because, at least for Camden and Somerset, substantial

disparities also exist between jury pools with the broader community. In other

words, the drop-off between the venire and the jury compounds a pre-existing

problem in African Americans’ levels of representativeness for these two areas.28

26 Of note, the comparative disparities appearing in parentheticals are not the same as

statistical tests. As I explain in Appendix B, because jury service is a fairly low-

probability event for anyone, differences in participation rates must be very large to test

as statistically significant, which renders it a highly conservative test. Appendix B

reviews what statistical testing entails conceptually, which statistical models I used, and

results from those tests.

27 Additionally, Morris County had no African American jurors, but it also began with

venires that were just 1% African American, and this was fairly representative of the true

Morris County population (see Table III.1).

28 It bears mentioning that the Somerset jury vs. venire comparison is based on just one

trial, which could have been atypical or had idiosyncratic practices that contributed to the

disparity in African American participation.

49

Table IV.1: Comparison of Venire and Jury Composition, For All Juries, by

County.

N Trials % in venire % on juries

County All Whites Blacks Latinos Asians Whites Blacks Latinos Asians

Bergen 13 70.7 4.8 15.2 12.9 77.0 7.0 (10.0) (7.0)

Burlington 8 76.3 16.1 4.0 3.0 82.3 (9.7) 8.1 (0)

Camden 8 76.4 12.6 9.1 4.0 84.6 (5.8) 9.6 (1.9)

Cumberland 2 68.0 14.0 18.6 1.7 65.3 23.1 15.4 3.9

Essex 12 48.2 31.7 16.8 7.5 41.3 38.7 17.3 6.7

Gloucester 3 88.3 5.9 2.3 2.0 76.9 7.7 3.9 3.9

Mercer 5 67.4 12.0 6.8 13.8 61.2 18.4 6.1 12.2

Middlesex 9 60.9 9.4 15.0 21.3 63.3 10.0 16.7 21.7

Monmouth 9 84.0 2.7 9.6 8.0 84.1 4.6 13.6 6.8

Morris 3 81.7 1.0 8.9 11.2 71.4 (0) 28.6 (4.8)

Ocean 5 89.7 3.2 9.3 1.0 90.9 6.1 (3.0) (0)

Passaic 9 65.8 9.7 23.1 7.3 67.4 11.6 19.8 5.8

Somerset 1 73.0 4.9 5.7 17.2 85.7 (0) 14.3 (7.1)

Union 8 54.9 18.4 17.8 8.0 53.4 19.2 15.1 8.2

Note: Values in parenthesis indicate an underrepresentation of a given group on juries compared

to its percentage on the venire by at least a 25% comparative disparity.

At the same time, the starkest and most counter-intuitive finding in

Table IV.1 is that the majority of counties did not underrepresent Blacks on

juries. Fully ten of the fourteen counties studied represented or over-represented

African Americans on juries.

Turning to results for Latinos, six counties exhibited some level of

underrepresentation on juries, but Bergen and Ocean Counties saw the most

sizeable amounts: -5.2% and -6.3% respectively, which reflect substantial

50

comparative disparities (-34% and -68%). Four counties (Cumberland, Mercer,

Passaic, and Union) exhibited underrepresentation with disparities of 3.8% or less,

with low comparative disparities that ranged from about 10 – 17%. The other areas

saw no underrepresentation of Latinos on juries compared to venires. Six counties

also had underrepresentation of Asians, but for three of these (Burlington, Camden,

and Ocean), the area had few Asian Americans in its venires (less than 3%). By

contrast, Bergen, Morris, and the single trial in Somerset greatly underrepresented

Asians both in terms of absolute disparities (-5.9%, -7.8%, and -10.1%,

respectively) and comparative disparities (-46%, -70%, and -59%, respectively).

Hence, for Latinos and Asians, as for African Americans,

underrepresentation was sizeable and concerning in two or three counties;

however, in the majority of all areas studied, these groups were not starkly

underrepresented.

B. Criminal vs. civil juries. Although the overall analysis presented above,

which ignored case type, has the benefit of examining a large number of trials,

there are two reasons to look at patterns of underrepresentation across criminal

versus civil trials. First, because African Americans, in general, have a fraught

relationship with law enforcement, it would be particularly concerning if they were

reliably underrepresented on criminal cases in particular.

51

Table IV.2: Comparison of Venire and Jury Composition, For Criminal Cases,

by County.

N Trials % in Criminal Trial Venires % on Criminal Juries

County Criminal Whites Blacks Latinos Asians Whites Blacks Latinos Asians

Bergen 2 70.9 4.3 17.7 11.2 81.5 3.7 14.8 (0)

Burlington 2 74.8 17.2 4.0 3.0 78.3 (13.0)* 8.7 (0)

Camden 1 76.5 15.0 8.0 2.7 55.6 33.3 11.1 (0)

Cumberland 2 68.0 14.0 18.6 1.7 65.4 23.1 15.4 3.9

Essex 5 46.0 33.4 17.4 7.8 38.1 40.5 19.1 7.1

Gloucester 1 89.0 4.9 1.2 1.8 69.2 15.4 (0) (0)

Mercer 3 67.5 12.1 7.6 12.8 63.6 18.2 9.1 (6.1)

Middlesex 1 60.0 6.0 16.0 20.0 58.3 16.7 (8.3) 25.0

Monmouth 1 86.3 4.1 6.9 5.5 81.8 (0) 18.2 9.1

Morris 0 -- -- -- -- -- -- -- --

Ocean 1 84.6 4.6 10.0 1.8 88.9 11.1 (0) (0)

Passaic 3 64.5 8.9 23.4 8.1 69.8 (4.7) 20.9 7.0

Somerset 1 73.0 4.9 5.7 17.2 85.7 (0) 14.3 (7.1)

Union 3 53.2 21.8 17.8 8.2 53.7 19.5 17.1 7.3

Note: Values in parenthesis indicate an underrepresentation of a given group on juries compared

to its percentage on the venire by at least a 25% comparative disparity. (*This disparity is 24%.)

The second reason hypothesizes an opposite pattern. In New Jersey criminal

juries are larger in size than civil juries. Thus, criminal juries should be more likely

achieve greater diversity, and there should be lower levels of underrepresentation

on criminal rather than civil juries. Table IV.2 presents patterns of representation

in the 26 criminal trials in the data, whereas results for the 69 civil trials are in

Table IV.3. In reviewing criminal trial results, it is important to note that several

areas contributed just one criminal trial to the study, which can lead to results –

52

both in terms of levels of overrepresentation or underrepresentation – that may

have been idiosyncratic to a particular case.

Compared to the overall analysis, a few more counties profile as having

substantial underrepresentation when data are split by trial-type (criminal vs. civil).

Thus, patterns of underrepresentation described in the overall analysis are

not typically constant across the two types of trials. For African Americans,

Monmouth and Passaic had comparative disparities that exceeded 25% for criminal

cases, although Monmouth’s result stemmed from just one trial; Burlington

(discussed in the combined analysis) had a comparative disparity just shy of 25%.

On the civil side, in addition to Burlington underrepresenting African Americans,

two particular results in Table IV.3 bear mentioning. Two civil trials in Gloucester

during the study period seated no African Americans on juries despite Blacks

being, on average, over 11% of venires; in Camden, not a single one of this

county’s seven civil cases seated an African American juror. In both of these

counties, African Americans were overrepresented in each areas’ lone criminal

case. These are clear examples of criminal juries doing a substantially better

job representing African Americans than civil cases.

53

Table IV.3: Comparison of Venire and Jury Composition, For Civil Cases, by

County.

N % in Civil Trial Venires % on Civil Juries

County Civil Whites Blacks Latinos Asians Whites Blacks Latinos Asians

Bergen 11 70.4 5.2 12.6 14.6 75.3 8.2 (8.2) (9.6)

Burlington 6 77.1 15.6 3.9 2.9 84.6 (7.7) 7.7 (0)

Camden 7 76.4 11.4 9.6 4.7 90.7 (0) 9.3 (2.3)

Cumberland 0 -- -- -- -- -- -- -- --

Essex 7 51.7 28.9 15.8 7.1 45.5 36.3 15.2 6.1

Gloucester 2 87.1 7.5 4.3 2.2 84.6 (0) 7.7 7.7

Mercer 2 66.1 11.3 1.6 21.0 56.3 18.8 (0) 25.0

Middlesex 8 61.1 9.9 14.9 21.6 64.6 8.3 18.9 20.8

Monmouth 8 83.6 2.4 10.1 8.5 84.6 6.1 12.1 (6.1)

Morris 3 81.7 1.0 8.9 11.2 71.4 (0) 28.6 (4.8)

Ocean 4 93.0 2.3 8.8 1.0 91.7 4.2 (4.2) (0)

Passaic 6 68.3 11.4 22.5 5.5 65.1 18.6 18.6 4.7

Somerset 0 -- -- -- -- -- -- -- --

Union 5 57.1 13.8 18.0 7.8 53.3 18.8 (12.5) 9.4

Note: Values in parenthesis indicate an underrepresentation of a given group on juries

compared to its percentage on the venire by at least a 25% comparative disparity.

Latinos also had somewhat better representation on criminal rather than civil

juries; however, patterns were more mixed and never as clear-cut as the instance

just described. On the one hand, some counties had better representation on civil

rather than criminal cases. Middlesex County’s single criminal trial

underrepresented Latinos by 7.7% (comparative disparity of 48%), and

Gloucester’s lone criminal case seated no Latino member; however, in this latter

instance, Latinos made up less than 2% of the venire. By contrast, a few counties

54

saw superior representation on criminal juries. In the overall analysis, Union

County did not reach a 25% comparative disparity threshold, but when cases were

limited to just civil trials, the underrepresentation level (-5.5%) generated a

comparative disparity of 31%. In Bergen and Mercer Counties, there was more

substantial underrepresentation on Latinos in civil trials than in criminal cases, but,

again, Mercer had a low prevalence of Latinos in the venire. Only Ocean County

had concerning underrepresentation in both civil and criminal trials.

Looking at results for Asian-American jurors, four counties saw

underrepresentation in both criminal and civil cases: Bergen, Burlington, Camden,

and Ocean. However, consistent with the discussion of the overall results, this is

linked to Asian-American’s limited presence in the venire. Of the four counties

with 25% comparative disparities across both trial types, in only Bergen County

were Asian Americans represented in venires at a non-trivial level (i.e., about 11 –

15% across the two trial types). Mercer and Gloucester had underrepresentation in

criminal but not civil trials, whereas the reverse occurred in Monmouth.

Zero group members? In advocating for larger-sized juries (e.g., 12 instead

of 6), jury scholars have said that the larger groups improve representation not only

or even necessarily by changing the average representation levels.29 They suggest

29 See, e.g., Shari Seidman Diamond, Destiny Peery, Francis J. Dolan, and Emily Dolan,

Achieving diversity on the jury: Jury size and peremptory challenges, 6 JOURNAL OF

EMPIRICAL LEGAL STUDIES 425 (2010); Michael J. Saks, JURY VERDICTS (1977).

55

that smaller groups are particularly prone to wider swings in participation rates of

minority group members – smaller groups are more likely, for example, to have

none or a lot of a given group. This reflects the statistical reality that the smaller

the group, the more variability the group exhibits, including variability in the

profile of the people selected for a smaller-sized group.30 I therefore supplemented

the above analysis of rates of participation across civil and criminal juries with a

basic count: For each of the three groups examined, how often did jury selection

end with no members of that group on a jury? For this analysis, I counted only

instances in which there was at least one member of the group in a venire.

Jury size is linked with the likelihood that a jury will contain no

members of a racial/ethnic group. For African Americans, there were 33

instances of zero African Americans on a jury across 85 trials that included at least

one Black venire-member; that is, 39% of cases that could have seated a Black

juror had none. The rate of juries without any African American members was,

however, significantly lower on criminal cases, 5 of 26 or 19%, compared to civil

cases, 28 of 59 or 47%.31 The pattern is the same for Latinos. There were 36

30 Id.

31 This result comes from a statistical test called a “chi-square test of association.” This

test produced a chi-square value of 6.05 (df = 1, N = 85), p < .05. Results are nearly

identical and significant if I remove the three counties that had only one type of trial or

another.

56

instances of juries lacking a Latino/a member across 91 trials with at least one

Latino/a venire-member, a rate of 40%. This outcome was more likely to occur on

civil trials (48%) than criminal trials (29%).32 Unlike the other two groups, it was

more common than not for a jury to have no Asian-American members (58% of

trials), and the difference between criminal cases (50%) and civil cases (61%) did

not test as statistically significant.33

To sum up this sub-section on representation levels on criminal versus civil

cases: except for analyses of Asians, a group that had low prevalence in several

areas, typically groups were not underrepresented on both civil and criminal

cases. Such consistency was true only for Burlington County and African

Americans, and Ocean County with respect to Latinos. I observed little evidence

that African Americans are particularly likely to be underrepresented on

criminal cases. By contrast there was some evidence that civil cases are more

likely to underrepresent a group compared to criminal cases. In three counties

Latinos enjoyed greater representation on criminal rather than civil juries, whereas

two counties showed the reverse. The most striking pattern occurred for African

32 This is again a statistically significant difference: chi-square = 6.29 (df = 1, N = 91), p

< .05, and results do not change markedly if I eliminate the three counties that had just

one type of trial or the other.

33 Chi-square = 0.91 (df = 1, N = 90), p < .35.

57

Americans in two counties, each with non-trivial proportions of African Americans

in their venires: African Americans were overrepresented on each county’s lone

criminal trial, whereas multiple civil cases seated no African Americans. This

result was particularly dramatic in one county, which held fully seven separate

civil cases with the same pattern. Finally, analyses indicate that civil juries are

more likely than larger-sized criminal juries to fail to have any African American

or Latino members.

C. Conclusions and Recommendations #4 and #5. The first conclusion from

analysis of minority representation across fourteen counties is a surprising one: In

many ways, New Jersey courts are doing an admirable job ensuring that

minority group members in a venire participate on criminal and civil juries.

In the majority of counties, representation levels of African Americans, Latinos,

and Asian Americans on juries resembled or exceeded the group’s prevalence in

the venires. As detailed in Appendix B, there were also few statistical differences

in rates of participation on juries for minority-group members compared to

White/non-Hispanic venire-members. Underrepresentation was not wholly absent,

but only two or three counties exceeded a 25% threshold of underrepresentation

within analyses of each minority group, and these were typically not the same

counties (i.e., counties did not typically underrepresent all groups).

58

Further, patterns were not typically fixed across type of case. Among

counties with underrepresentation of either Blacks or Latinos, just a single county

exhibited substantial underrepresentation in both civil and criminal trials

(Burlington for African Americans, Ocean for Latinos). Consistency in

underrepresentation across case types was a more common pattern in analyses of

Asian Americans; however, low prevalence in many counties appeared to account

for some, but not all, of this. Instead it was more common for a county to have

some underrepresentation in one type of case but not another.

Finally, there are several indicators that criminal juries, which are larger

in size, do a better job of representing minority group members than smaller-

sized civil juries. In two counties, Gloucester and Camden, no African Americans

appeared on any civil juries. I more formally tested whether civil juries were

disproportionately likely to have extreme participation patterns, specifically the

total absence of group members, compared to criminal trials. Results for Latinos

and African Americans supported findings from prior jury research studies: civil

cases were statistically more likely than criminal cases to have juries that failed to

seat any African Americans or any Latinos.

Recommendation #4: New Jersey courts should continue their

analysis of the practices that contribute to representative juries being

drawn from venires. Precisely because the juries in these fourteen

areas commonly profiled well on indicators of representativeness,

New Jersey should better understand what exactly contributed to this

result in multiple counties. As the study also points to particular areas

59

that have more consistent underrepresentation patterns, either across

case type or within case type, these areas should conduct a self-study

to assess what may be happening in these cases. (Sections V, VI, and

VII begin some of this work.) This is particularly important in the

areas in which no African Americans appeared on the mass of civil

cases.

Recommendation #5: One way for New Jersey to show its

commitment to seating more representative juries would be to

increase the size of civil juries to 12 members, rather than six. A

recent article by Federal judges argues for the superiority of the 12-

person jury,34 and the data in this study amply demonstrate why

larger-sized groups do a better job than smaller-sized groups of

achieving stable levels of representativeness and avoiding the

wholesale elimination of minority-group members.

34 Patrick Higginbotham, Lee Rosenthal, and Steven Gensler, Better by the Dozen:

Bringing Back the Twelve-Person Civil Jury, 104 JUDICATURE 46 (2020).

60

SECTION V: PEREMPTORY CHALLENGES

The previous section described the subset of concerning patterns of

underrepresentation across counties and observed that, particularly in civil cases,

minority group members are sometimes wholly eliminated from jury participation.

This section focuses on one way that people are excluded from jury participation:

the exercise of peremptory challenges.

A. Peremptory challenges and minority underrepresentation. Because New

Jersey grants parties a large number of peremptory challenges,35 a natural question

arises about what role peremptory challenges play in instances of minority

underrepresentation in New Jersey courts. According to the survey data collected,

peremptory challenges play a small role in explaining jury selection patterns

for minority groups; its minimal effect on underrepresentation stems from its

constrained use. Table V.1 presents results on the average number of peremptory

challenges used in these trials on African Americans, the variability in use – called

the “standard deviation” of the mean, which appears in parentheses next to the

35 According to N.J.S.A. § 2B:23-13: parties have six strikes per party in civil cases; as

many as 20 peremptories for the defense for categories of serious felonies described in

the statute when a defendant is tried alone (10 if tried jointly), whereas the State gets 12

challenges in these cases for a defendant tried alone, but 6 for each 10 the defense

receives when defendants are tried jointly; and 10 per side in other types of criminal

indictments (fewer if the case is “to be tried by a jury from another county”). The large

number of peremptories, particularly in criminal cases, is atypical among states (see, e.g.,

“State of the States Survey,” http://www.ncsc-

jurystudies.org/__data/assets/pdf_file/0016/5623/soscompendiumfinal.pdf, at 8).

61

average36 – as well as the maximum number of strikes used by each side on

African Americans in any trial. Results are presented separately for criminal and

civil trials and include only trials in which there was at least one African American

in the venire. Tables V.2 and V.3 present these same results for Latinos and Asian

Americans, respectively.

1. African Americans. According to Table V.1, peremptory challenges offer

some, but only minimal, purchase in explaining instances of underrepresentation of

African Americans. These challenges were, quite simply, not exercised very often

on this group. Looking broadly, across the 85 trials with at least one African

American in the venire, peremptory challenges by either side removed just 52 of

the 761 African Americans in these cases, or 7%. According to Table V.1 in no

instance did the overall average level of peremptory use per trial exceed one;

within each county, across both trial types, it was unusual for the even the

maximum number of strikes used on African Americans to exceed one. In the

single instance in which the maximum value was as high as 3 peremptory

36 The standard deviation is a useful descriptive statistic because it summarizes how

much the use of peremptories varied across multiple trials. One standard deviation added

or subtracted from the average value represents the range of fully two-thirds of the

distribution of peremptory usage (although in several instances, there were so few trials

that the word “distribution” is somewhat inapt). When peremptory use is highly variable,

the standard deviation will be larger relative to the mean. Note that when counties held

only one trial, no standard deviation is reported because there was no variability in usage.

62

challenges (used by the State in a criminal trial in Essex County), I reviewed the

circumstances. There were 57 African Americans in the venire, and the proportion

of African Americans on the final jury mimicked their representation in the venire.

In the data, by far the most common number of strikes against African Americans

(i.e., the “modal value”) was zero.

Table V.1: Use of Peremptory Challenges on African Americans Across Fourteen

Counties: Means, Standard Deviations, and Maximum Number Used, for Criminal

and Civil Cases.

Criminal Trials Civil Trials

County Trials State Max Defense Max Trials Plaintiff Max Defense Max

Bergen 2 0 0 0 0 8 0.1 (0.4) 1 0.3 (0.5) 1

Burlington 2 1.0 (0.0) 1 1.0 (1.4) 2 6 0.2 (0.4) 1 0.7 (0.5) 1

Camden 1 1.0 ( -- ) 1 0 0 7 0.1 (0.4) 1 0.1 (0.4) 1

Cumberland 2 0.5 (0.7) 1 0.5 (0.7) 1 0 -- -- -- --

Essex 5 0.6 (1.3) 3 0.2 (0.4) 1 7 0.1 (0.4) 1 0.4 (0.5) 1

Gloucester 1 0 0 0 0 2 0.5 (0.7) 1 1.0 (1.4) 2

Mercer 3 0.7 (0.6) 1 0 0 2 0 0 0 0

Middlesex 1 0 0 0 0 7 1.0 (0.8) 2 0.1 (0.4) 1

Monmouth 1 1.0 (--) 1 0 0 6 0.2 (0.4) 1 0 0

Morris 0 -- -- -- -- 1 0 0 0 0

Ocean 1 0 0 0 0 2 0 0 0 0

Passaic 3 0.3 (0.6) 1 0 0 6 0.3 (0.8) 2 0.5 (0.5) 1

Somerset 1 0 0 0 0 0 -- -- -- --

Union 3 0.3 (0.6) 1 0.7 (0.6) 1 5 0.4 (0.5) 1 0.2 (0.4) 1

Overall 26 0.5 (0.7) 3 0.2 (0.5) 2 59 0.3 (0.6) 2 0.3 (0.5) 2

Note: “( -- )” indicates instances of just one trial which does not permit a standard deviation estimate.

Results based on the 85 trials that had at least one African American in the venire.

63

Of course, even infrequent use of peremptories can have effects. A

significant concern about peremptories is that they will “pick off” members of

communities that have a small presence in a venire – removing, for example, the

only two African Americans in a venire with just two peremptory strikes; studies

find even this level of elimination can have an effect on trial outcomes.37 I looked

for evidence of this type of problematic result in three ways. First, I looked for any

instance of this precise pattern. For African Americans, there was no instance in

which peremptories removed all available Black venire-members.

Second, using the same threshold I have employed throughout this report, I

counted the number of times that the combined use of peremptory challenges by

both sides removed at least 25% of the members present in the venire. For African

Americans, I counted 13 trials, across eight different counties, that met this

threshold. Across these trials, a total of 23 peremptories removed African

Americans, and these venires had a total of 59 African American venire-members;

thus, in this subset of 13 cases with high-impact peremptory use, peremptories

removed fully 39% of African Americans in those venires. This is a concerning

result, but it bears repeating that it unusual across all trials: these 13 trials

constituted just 15% of the 85 cases that had at least one African American in the

37 See, e.g., Shamena Anwar, Patrick Bayer, and Randi Hjalmarsson, The Impact of Race

in Criminal Trials, 127 QUARTERLY JOURNAL OF ECONOMICS 1017 (2012).

64

venire. Further, of the eight counties demonstrating this pattern of removing at

least 25% of African Americans in any one trial, only four (Burlington, Camden,

Gloucester, and Monmouth) were mentioned at all in Section IV as having

concerning levels of representation in either their civil or criminal juries. The

remaining counties did not exhibit underrepresentation on the trials studied,38 or

had underrepresentation but peremptories removed fewer than 25% of the African

Americans in the venires.

The final way I looked for potentially problematic use of peremptories was

to pull out the 33 cases, discussed in Section IV, in which no African American

made it onto the jury, even though there was at least one African American in the

venire. I looked at attrition in these cases and examined in how many instances the

peremptory contributed to the outcome of zero Black jurors. Of these 33 cases, just

under half (n = 16, 48%) involved an attorney using even one peremptory

challenge on African American venire-members.39 In cases in which no African

Americans made it to the jury and attorneys used a peremptory, peremptories

38 Passaic also had one instance in which peremptories removed at least 25% of the Black

venire-members, but this occurred in a civil case, whereas the underrepresentation

described in Section IV was on a criminal trial.

39 The remaining 17 cases seated no African American jurors because of heavy use of

challenges for cause or because the African American venire-members were listed as “not

used.” These sources of attrition are discussed in more detail below in Sections VI and

VII.

65

removed just over one quarter of the 83 African Americans in the venire from jury

service (n = 22, 27%). This result again demonstrates that peremptories played a

role in the wholesale removal of African Americans from these juries; however,

these strikes were by no means the primary culprit. For the majority of African

Americans, and in the majority of trials examined, attrition came through other

mechanisms. This conclusion is particularly apt for the unusual instance of

Camden, in which all seven of its civil trials failed to seat even one African

American. Across those seven cases, just 2 peremptory strikes in total were used

against African Americans (in two trials, one in each); thus, peremptories do not

explain patterns for this county.

2. Latinos. Analysis of Latinos produced similar broad patterns of

peremptory use as for African Americans, but attorneys used peremptories to strike

Latinos a bit more frequently than was true for Blacks. For example, as with

African Americans, the modal value for peremptory usage on Latinos in both civil

and criminal trials was zero, and as Table V.2 shows, the most common average

value for peremptory use was 1.0 or lower. At the same time, the proportion of

Latinos removed via the peremptory was somewhat higher: attorneys struck a total

88 Latino individuals, or 10% of the 879 Latinos in venires. Further, Table V.2

shows higher maximum values occurred somewhat more frequently. There were

five cases in which attorneys for one side or the other used as many as 3

66

peremptories on Latino venire-members, and in one unusual case, the defense used

8 (this was an outlier case from Passaic, in which the defense in a civil case used a

total of 29 peremptory strikes; no other case had as many as twenty strikes used by

a side, and in this instance, Latinos still made up 25% of the jury).

Table V.2: Use of Peremptory Challenges on Latinos Across Fourteen Counties:

Means, Standard Deviations, and Maximum Number Used, for Criminal and Civil

Cases.

Criminal Trials Civil Trials

County Trials State Max Defense Max Trials Plaintiff Max Defense Max

Bergen 2 1.0 (1.4) 2 0 0 11 0.3 (0.6) 2 0.3 (0.5) 1

Burlington 2 0 0 0 0 5 0 0 0.4 (0.5) 1

Camden 1 0 0 0 0 7 0.1 (0.4) 1 0.1 (0.4) 1

Cumberland 2 1.0 (0.0) 1 2.0 (1.4) 3 0 -- -- -- --

Essex 5 0.6 (0.9) 2 0.2 (0.4) 1 7 0 0 0.3 (0.8) 2

Gloucester 1 0 0 0 0 2 0 0 0 0

Mercer 3 0 0 1.0 (0.0) 1 1 0 0 0 0

Middlesex 1 1.0 (0.0) 1 0 0 7 0.7 (0.8) 2 0.4 (0.5) 1

Monmouth 1 1.0 ( -- ) 1 0 0 7 0.1 (0.4) 1 0.4 (0.8) 2

Morris 0 -- -- -- -- 3 0.3 (0.6) 1 0.7 (0.6) 1

Ocean 1 3.0 ( -- ) 3 1.0 ( -- ) 1 4 0.3 (0.5) 1 0.5 (0.6) 1

Passaic 3 1.7 (1.5) 3 2.0 (1.7) 3 6 0.7 (1.2) 3 2.3 (3.1) 8

Somerset 1 0 0 0 0 0 -- -- -- --

Union 3 0.3 (0.6) 1 0.7 (0.6) 1 5 0.4 (0.9) 2 0.4 (0.5) 1

Overall 26 0.7 (1.0) 3 0.7 (1.0) 3 65 0.3 (0.7) 3 0.5 (1.2) 8

Note: “( -- )” indicates instances of just one trial which does not permit a standard deviation estimate.

Results based on the 91 trials that had at least one Latino/a in the venire.

Taking a more granular look at all the trials, I again investigated whether

even a small number of peremptories undermined the representation of Latinos on

67

New Jersey juries. Findings are consistent with the gist of the broader look:

although peremptories played a mostly small role in contributing to

underrepresentation, effects of the peremptory on Latinos were a bit stronger than

for African Americans. First, there was one instance of a civil trial in Burlington in

which one peremptory challenge eliminated the only Latino from the venire.

Second, there were more (n = 18) instances in which peremptories removed at least

25% of Latinos in the venire.40 Those 18 cases represent 20% of the 91 trials with

at least one Latino/a venire-member, a somewhat higher level than that observed

for African Americans; it also occurred in more counties (n = 11 compared to 8 for

African Americans.)

The link to final underrepresentation on juries through these practices

nonetheless remains tenuous: just three of the 11 counties in which peremptories

eliminated at least 25% of the Latino members of the venire also profiled in

Section IV as having concerning underrepresentation on juries (Bergen, Ocean,

and Union). Finally, I carefully examined the 36 cases in which no Latino/a ended

up on the jury. Here the pattern was very similar to that for African Americans: in

only 16 cases (44%) were any peremptories used; they removed 20% of the 127

40 Interestingly, only one of these cases was the same trial in which attorneys used

peremptories against African Americans in ways that removed at least one-quarter of

their numbers on the venire.

68

Latinos in these venires, a non-negligible amount but by no means constituting a

majority.

3. Asian Americans. Given that Asian Americans were frequently just a

small portion of venires (<5% - see Tables IV.1 to IV.3), peremptories would seem

to pose a substantial risk to the presence of this group on juries. However,

according to Table V.3, as was with other groups, peremptory strikes were used

infrequently on this group. A total of 567 Asians appeared in the venires of these

trials, and attorneys exercised 44 peremptories against this group (8%). In just one

case was the lone Asian American venireperson eliminated through a peremptory.

In 13 trials from seven counties, peremptories removed at least 25% of the Asian

Americans in the venire; however, just three of these counties – Bergen, Camden,

and Gloucester – exhibited concerning levels of underrepresentation in the analysis

presented in Section IV, again pointing to the likelihood that other factors

determine how representative the jury will be. Among the 52 cases in which no

Asian Americans served on juries despite there being at least one Asian American

in the venire, peremptories played no role in in fully 35 of them (67%) because no

peremptories were exercised on Asian Americans during these trials. In the

remaining 17 trials, a bare majority (9 of 17) saw the loss of Asian Americans in

the venire exceed 25%.

69

Table V.3: Use of Peremptory Challenges on Asian Americans Across Fourteen

Counties: Means, Standard Deviations, and Maximum Number Used, for Criminal

and Civil Cases.

Criminal Trials Civil Trials

County Trials State Max Defense Max Trials Plaintiff Max Defense Max

Bergen 2 0.5 (0.7) 1 1.5 (2.1) 3 11 0.3 (0.6) 2 0.3 (0.5) 2

Burlington 2 0 0 0 0 3 0 0 0 0

Camden 1 0 0 0 0 7 0.3 (0.5) 1 0.1 (0.4) 1

Cumberland 2 0 0 0 0 0 -- -- -- --

Essex 5 0 0 0 0 7 0.1 (0.4) 1 0.1 (0.4) 1

Gloucester 1 0 0 1.0 ( -- ) 1 2 0 0 0 0

Mercer 3 0 0 0.3 (0.6) 1 2 0 0 0.5 (0.7) 1

Middlesex 1 1.0 ( -- ) 1 0 0 8 0.6 (0.7) 2 0.8 (0.9) 2

Monmouth 1 0 0 1.0 ( -- ) 1 8 0.1(0.4) 1 0.1 (0.4) 1

Morris 0 -- -- -- -- 3 1.0 (0.0) 1 0 0

Ocean 1 0 0 0 0 3 0 0 0 0

Passaic 3 1.0 (1.0) 2 0.3 (0.6) 1 5 0.2 (0.4) 1 0.2 (0.4) 1

Somerset 1 0 0 0 0 0 -- -- -- --

Union 3 0.3 (0.6) 1 0 0 5 0 0 0.2 (0.4) 1

Overall 26 0.2 (0.5) 2 0.3 (0.7) 3 64 0.3 (0.5) 2 0.2 (.5) 2

Note: “( -- )” indicates instances of just one trial which does not permit a standard deviation estimate.

Results based on the 90 trials that had at least one Latino/a in the venire.

4. Summary. Although there were small variations in patterns across

analyses of African Americans, Latinos, and Asian Americans, a fair conclusion

for all groups is that peremptories play a role in underrepresenting these

groups on some juries, but only an attenuated one. According to a variety of

measures, peremptory challenges are rarely the primary way that minority

groups experience attrition from jury participation.

70

B. Peremptory use overall. The infrequent use of peremptory challenges on

the groups analyzed above is noteworthy given how many challenges New Jersey

allots to parties. Is such infrequency specific to minority group members? Perhaps,

contrary to most scholarship in this domain, all parties are disproportionately

aiming strikes at the non-Hispanic White members of a venire, leaving few

available to exercise on minority group members. One way to consider this

possibility is to run statistical tests that ask whether White/non-Hispanic

venirepersons are actually more likely to be dismissed through a peremptory

challenge than are members of the three groups I examined above. I did so, testing

for effects due to race/ethnicity while also accounting for variability in peremptory

use across trials (see Appendix B for more discussion of this approach as applied

to jury participation). I tested models for criminal and civil trials separately, and

also separately for each side (i.e., for the state/plaintiff’s peremptory challenges

and for the defense’s use of strikes).41

41 It is particularly important to separate out case type because research has shown that

although it is the plaintiff (the State) in criminal trials who is more likely to dismiss

African Americans than to dismiss Whites, it is the defendant in civil trials who

disproportionately dismisses African Americans rather than White venirepersons (cf.

Diamond et al., supra, note 29; Mary R. Rose, The Peremptory Challenge Accused of

Race or Gender Discrimination? Some Data from One County, 23 LAW & HUMAN

BEHAVIOR 695, 1999).

71

Only one model produced a significant effect: Criminal defense attorneys

were about one-fourth less likely to dismiss African Americans through a

peremptory than they were to use a peremptory strike on non-Hispanic Whites (p <

.01). Of the 135 peremptory challenges exercised by criminal defense attorneys,

just six were against African American venirepersons, whereas 114 were against

Whites. Otherwise, rates in peremptory use on non-Hispanic Whites versus the

other groups in all other tests were statistically comparable.42

How often do attorneys exercise strikes on anyone? Table V.4 on the next

page presents data on the average number, standard deviation, and maximum value

of peremptory challenges used on any venireperson across all counties, by county

and by attorney type. In the bottom rows, I present additional information about the

distribution of peremptory challenges in New Jersey by indicating the number of

peremptories used by each side when that side was at the 75th and 90th percentile of

the distribution. Put another way, these figures indicate what amount of

42 As Appendix B describes, statistical testing is a conservative approach to identifying

differences in peremptory use across racial/ethnic groups because being dismissed by a

peremptory in these data is a rare event for everyone: fewer than 10% of venirepersons

were eliminated through the peremptory. In such situations, effects will not test as

significant unless the differences are sizeable, as was the finding for defense attorneys’

use of peremptories on Black versus White venirepersons. Thus, results indicate no other

instance of a difference this extreme.

72

peremptory strikes occurred in the top 25% and 10%, respectively, of cases that

used peremptories.

Table V.4: Use of Peremptory Challenges Across Fourteen Counties: Means, Standard Deviations, and Maximum Number Used, for Criminal and Civil Cases.

Criminal Trials Civil Trials

County Trials State Max Defense Max Trials Plaintiff Max Defense Max

Bergen 2 3.5 (4.9) 7 4.5 (6.4) 9 11 1.8 (1.8) 5 2.4 (3.2) 11

Burlington 2 1.5 (0.7) 2 3.0 (0.0) 3 6 2.8 (1.8) 5 2.5 (1.9) 6

Camden 1 2.0 ( -- ) 2 2.0 ( -- ) 2 7 3.0 (0.8) 4 2.6 (1.8) 6

Cumberland 2 3.0 (1.4) 4 11.5 (11) 19 0 -- -- -- --

Essex 5 3.2 (3.6) 9 4.6 (4.9) 12 7 2.3 (1.8) 5 1.7 (1.4) 4

Gloucester 1 7.0 ( -- ) 7 13.0 ( -- ) 13 2 3.5 (0.7) 4 4.5 (0.7) 5

Mercer 3 4.3 (2.1) 5 4.3 (1.2) 5 2 1.5 (2.1) 3 1.0 (1.4) 2

Middlesex 1 3.0 ( -- ) 3 2.0 ( -- ) 2 8 3.8 (1.7) 6 3.1 (2.2) 6

Monmouth 1 7.0 ( -- ) 7 12.0 ( -- ) 12 8 1.9 (1.4) 4 3.3 (3.3) 11

Morris 0 -- -- -- -- 3 4.3 (1.5) 6 2.3 (0.6) 3

Ocean 1 8.0 ( -- ) 8 5.0 ( -- ) 5 4 3.5 (1.3) 5 3.5 (1.7) 5

Passaic 3 6.3 (0.6) 7 9.3 (4.0) 13 6 3.5 (5.7) 15 8.0 (10.8) 29

Somerset 1 4.0 ( -- ) 4 6.0 ( -- ) 6 0 -- -- -- --

Union 3 2.7 (2.1) 5 3.7 (2.3) 5 5 2.4 (1.1) 4 2.0 (1.2) 3

Overall 26 3.8 (2.6) 9 5.9 (4.8) 19 69 2.7 (2.2) 15 3.1 (3.9) 29

75th percentile -- 6 9 -- 4 4

90th percentile -- 7 13 -- 5 6

Note: “( -- )” indicates instances of just one trial which does not permit a standard deviation estimate.

Table V.4 makes clear that attorneys rarely use the full complement of

strikes allotted to them under statute. In criminal cases, the prosecution used, on

average, just under four strikes; those in the top 25% of the distribution used six,

and those in the top 10% used seven. Stated in terms of number of cases, in all but

73

six of the 26 criminal trials, prosecutors used fewer than seven peremptory

challenges (all but one used eight or fewer). For criminal defense attorneys, who

have more peremptory challenges allotted to them and therefore used more on

average, they nonetheless also did not use all their strikes. The top 75% of the

distribution across cases used nine or more strikes; the top 10% used 13. Stated in

terms of cases, all but seven cases used 8 or fewer peremptory strikes and all but

five trials used 10 or fewer.

In civil cases, each side had six strikes per party; clearly some trials had

multiple plaintiffs or defendants,43 since the maximum value in a few cases was

above ten and reached as high as 29. Despite these unusual values, the bottom of

Table V.4 shows that, on average, each side used around three strikes. The top

25% of attorneys on each side used four strikes, and the top 10% of civil plaintiffs

used five, whereas civil defendants used six. Stating results in terms of the raw

number of the 69 civil cases examined in this study, civil plaintiffs in all but four

cases used 5 or fewer peremptory strikes, and all but one used 6 or fewer. For civil

defendants, all but four used 6 or fewer. In other words, among civil cases, parties

in 65 of the 69 trials did not use even the six strikes allotted to them.

43 The data provided did not give information about number of parties, or any other case-

related information.

74

C. Conclusions. This section presented a series of surprising,

counterintuitive findings. First attorneys’ use of peremptory strikes on minority

group members played a case-specific and generally attenuated role in explaining

patterns of underrepresentation on juries. For example, in just under half of the

cases in which either African Americans or Latinos were wholly eliminated from

juries did attorneys use any peremptory challenges, and just a few of the counties

that profiled as having concerning levels of underrepresentation on civil or

criminal cases (see Section IV) also had concerning levels of peremptory usage.

Second, with one exception regarding criminal defendants’ use of peremptories on

African American compared to non-Hispanic White jurors, there is little strong

statistical evidence that peremptory challenge outcomes differed substantially by

race or ethnicity in these trials. Finally, an examination of peremptory challenge

use – both within each of the minority groups and across the entire sample – shows

that attorneys do not use a substantial number of challenges, and only rarely use all

or close to their full complement of strikes.

Given the largely restrained use of peremptory strikes in a state with such a

generous allotment, it is tempting to look at the data and conclude that attorneys,

particularly attorneys in criminal cases, do not need as many strikes as they have.

As an empirical matter, the data indicate that prosecutors in all but the rarest of

cases would need 8 strikes, and likewise the vast majority of criminal defendants

75

would need but 10; on the civil side, six strikes in total would have covered all but

a handful of parties in these trials. Because I conclude that peremptories have some

role, albeit a small one, in affecting jury selection outcomes, New Jersey may wish

to further limit the number of challenges attorneys use. However, the peremptory’s

effect on representativeness is, in these data, a fairly weak ground on which to

argue for limiting the number of peremptories. I present evidence for a more

clearly problematic effect of having so many peremptory challenges in Section

VII, which reviews the large venire sizes in New Jersey trials, particularly in

criminal cases.

As New Jersey works on deciding the “right” number of peremptory

challenges to permit, I do not recommend relying solely on the empirical account

of how attorneys used their strikes as presented in this Section. As I next discuss,

understanding peremptory use in these cases requires a careful look at another

source of attrition from jury service: judges’ generous use of challenges for cause.

76

SECTION VI. CHALLENGES FOR CAUSE

In the trials I studied, a total of 7,407 individuals underwent a voir dire

experience in either a criminal or civil trial. In criminal cases, the most common

way for individuals to conclude their voir dire experience was through a

challenge for cause; in civil cases, challenges for cause were a close second to

the group that was “not used.” In criminal cases, in which the slight majority of

the sample (3,753, or 51%) participated, the JMS database lists fully 57% of

people as having been dismissed for cause, which dwarfed the frequency of all

other categories, including the 26.7% of criminal venirepersons who were listed as

“not used.” By contrast, in civil cases, which had smaller venires on average and

included 3,654 venirepersons, the “not used” category was slightly larger (39.5%)

than the cause-challenge category (36.5%). Across all cases, nearly 47% of people

ended jury selection by being dismissed through a challenge for cause. The bottom

row of Table VI.1 presents these figures, with the bulk of the table depicting

specific patterns across counties. Given the high rates of people being dismissed

through challenges for cause, this section explores two questions: What

relationship do challenges for cause have with the use of peremptory challenges,

and is their use negatively affecting jury diversity?

77

Table VI.1: Voir Dire Outcomes by County and Trial Type.

All Type of Outcome for All

Venirepersons (%)

Type of Outcome for All

Criminal Venirepersons (%)

Type of Outcome for Civil

Venirepersons (%)

County N

People

Jurors Cause Pros/

Pltf

Def P Not

Used

N Juro

rs

Caus

e

Pros/

Pltf

Def P Not

Used

N Juror

s

Caus

e

Pros/

Pltf

Def P Not

Used

Bergen 1,092 9.5 51.6 2.3 3.2 33.2 551 5.3 70.6 1.3 1.6 21.2 541 13.9 32.2 3.7 4.8 45.5

Burlington 333 20.4 21.3 6.0 6.3 45.9 123 22.8 26.8 2.4 4.9 43.1 210 19.0 18.1 8.1 7.1 47.6

Camden 598 9.7 18.9 3.8 3.3 64.2 194 6.7 23.7 1.0 1.0 67.5 404 11.1 16.6 5.2 4.4 62.6

Cumberland 177 15.3 49.2 3.4 13.0 19.2 177 15.3 49.2 3.4 13.0 19.2 0 -- -- -- -- --

Essex 1,092 10.4 54.6 2.9 3.2 28.8 654 10.1 51.7 2.4 3.5 32.3 438 11.0 58.9 3.7 2.7 23.7

Gloucester 277 10.1 41.9 5.1 7.9 35.0 175 8.0 52.6 4.0 7.4 28.0 102 13.7 23.5 6.9 8.8 47.1

Mercer 545 10.3 67.5 2.4 2.8 17.1 478 8.4 68.2 2.1 2.7 18.6 67 23.9 62.7 4.5 3.0 6.0

Middlesex 433 15.7 47.3 7.6 6.2 23.1 60 23.3 51.7 5.0 3.3 16.7 373 14.5 46.6 8.0 6.7 24.1

Monmouth 665 10.5 50.2 3.3 5.7 30.2 91 14.3 42.9 7.7 13.2 22.0 574 9.9 51.4 2.6 4.5 31.5

Morris 175 12.0 22.9 7.4 4.0 53.7 0 -- -- -- -- -- 175 12.0 22.9 7.4 4.0 53.7

Ocean 330 12.4 19.4 6.7 5.8 55.8 133 10.5 18.8 6.0 3.8 60.9 197 13.7 19.8 7.1 7.1 52.3

Passaic 1,016 8.7 56.8 3.9 7.5 23.1 678 6.5 70.2 2.8 4.1 16.4 338 13.0 29.9 6.2 14.2 36.7

Somerset 125 11.2 44.0 3.2 4.8 36.8 125 11.2 44.0 3.2 4.8 36.8 0 -- -- -- -- --

Union 549 14.4 51.4 3.6 3.8 26.8 314 14.0 64.0 2.5 3.5 15.9 235 14.9 34.5 5.1 4.3 41.3

Overall 7,407 11.3 46.9 3.9 4.9 33.0 3,753 9.6 57.0 2.7 4.1 26.7 3,654 13.0 36.5 5.2 5.8 39.5

78

A. Cause challenges and peremptory challenges. A typical argument that

attorneys and scholars, including this author, 54 give for the necessity of

peremptory challenges is their role in offsetting judges’ unwillingness to dismiss

people who exhibit bias, particularly when prospective jurors verbally insist that

they can be fair. Under this argument, one would expect to see greater use of

peremptory challenges when judges excuse people for cause at lower rates.

Statistically, this is called a “negative relationship” or a “negative correlation:” as

one variable (cause challenge use) decreases, the other variable (peremptory

challenge use) increases. Given that cause challenges play such a common –

indeed, often the predominant – role in jury selection in New Jersey courts, the

notion of “stingy” judges who refuse challenges for cause seemed unlikely. But the

question nonetheless demands empirical testing.

To examine whether there is a negative relationship between for-cause and

peremptory challenges, I calculated the proportion of cause challenges used on

each venire – that is, the type of proportion depicted in Table VI.1, but at the trial-

level instead of the county-level. I also calculated the raw number of challenges

used by each side in each case and then tested for correlations between rates of

54 Mary R. Rose, and Shari Seidman Diamond. Judging Bias: Juror Confidence and

Judicial Rulings on Challenges for Cause, 42 LAW & SOCIETY REVIEW 513 (2008).

79

cause challenge grants in a venire and use of peremptories, examining civil and

criminal trials separately.

In New Jersey trials in this study, the purported negative relationship

between cause challenges and peremptory challenges does not exist. Instead,

in criminal cases, the higher the proportion of people who are dismissed

through a challenge for cause, the more challenges a defense attorney and a

prosecutor exercise. The correlation coefficients were .40 for the defense, and .41

for the state,55 both of which were statistically significant at p < .05. The

correlation coefficients for the civil cases did not test as statistically significant;

however, they were also not negative: .10 for the defendant’s use of peremptories

and a coefficient of zero for the plaintiff’s use. In short, there is no evidence that

attorneys use more strikes in reaction to judges’ being more conservative on

granting challenges for cause. The positive relationship between for-cause and

peremptory strikes signals that both judges and attorneys exercise challenges

in ways that are responsive the level of bias in the venire – as more people are

55 Correlations are bounded by -1.0 (two variables vary completely in tandem with one

another, but in a negative direction) to +1.0 (two variables vary completely in tandem

with one another, but in a positive direction, meaning that as one increases so does the

other). In between these two poles, a correlation coefficient of zero would indicate that

two variables are completely unrelated to one another.

80

excused for cause, attorneys also exercise more peremptory strikes, at least in

criminal cases.

I cannot overstate the importance of this finding for assessing whether or not

New Jersey has “too many” peremptory challenges, particularly if such

assessments are based on the observed (in)frequency of peremptory usage

described in Section V. Although attorneys rarely use their full complement of

peremptory challenges, in all likelihood, judges’ willingness to be generous in

excusing people for cause facilitates this outcome. Although I do not have any

data to prove why for-cause and peremptory strikes have a positive relationship

with one another in criminal cases, I would expect that attorneys are able to be

restrained in peremptory use when they do not have to use their strikes on

egregious cases of bias or on jurors who are, for example, deeply unhappy about

having to serve. They can instead be more selective in which people to strike. This

is advantageous given that a subset of excused jurors do take offense at being

excused through the peremptory challenge,56 and are less likely to feel that way

when excused by a judge. If New Jersey were to reduce the number of strikes it

makes available to attorneys, judges should not likewise pull back on granting

56 See, e.g., Mary R. Rose, A Voir Dire of Voir Dire: Listening to Jurors’ Views

Regarding the Peremptory Challenge, 78 CHICAGO-KENT LAW REVIEW 1061 (2003)

(reporting on interviews with excused prospective jurors).

81

challenges for cause. In all likelihood, the observed rates of peremptory use in

criminal trials depend upon judges’ willingness to use for-cause challenges.

B. Cause challenges and jury diversity. An understandable concern about

the high proportion of cause challenges in New Jersey venires is that liberal use

will negatively affect jury diversity. One might expect, for example, that economic

and other barriers to jury service are more common among non-Whites than among

Whites due to differences in personal resources that can address the burdens of

service (e.g., better-resourced people, who are more likely to be White, will have

advantages in absorbing childcare costs, reduced wages, and transportation costs

getting to and from service). Thus, to the extent that cause challenges are granted

on the basis of people expressing unwillingness to serve, or if minority-group

members are more alienated from courts or less willing to serve, then one would

expect cause challenges to disproportionately eliminate minorities from venires.

To assess whether there is a relationship between judges’ use of cause

challenges and jury outcomes, the first question to ask is whether the use of

challenges for cause is correlated with race or ethnicity. To investigate this, I ran

the same statistical models used to predict selection to a jury (Appendix B) and use

of peremptory challenges (see Section V). Specifically, I tested whether the

likelihood of being excused for cause was different for Blacks, Latinos, and Asians

than it was for non-Hispanic Whites. I ran models using both the aggregated data

82

(i.e., all areas combined) and also within each county. There was effectively no

relationship between a juror’s racial or ethnic background and the likelihood

that that person would be dismissed through a challenge for cause. The sole

exception was that Asian Americans were about one-and-a-half times more likely

to be dismissed by a cause challenge than Whites, but only in Bergen County –

hardly compelling evidence that judges are making decisions on the basis of

factors that disproportionately affect one racial or ethnic group more than others.

Given that judges do not disproportionately remove members of one group

or another, it seems unlikely that cause challenges are linked to jury composition.

But, to be certain, I tested for the presence of a statistically significant correlation

between the use of challenges for cause in a venire and that venire’s jury

composition outcome. To do this analysis, I again used the measure (discussed

above) that captures the proportion of challenges for cause used in a trial. I also

developed a measure of jury diversity, which I define as the proportion of non-

Hispanic Whites on each jury. To be sure, the latter is only a rough proxy for jury

diversity, since truly diverse juries are not defined by the how many members are

White, but rather by how well a jury matches the demographic diversity in the

venire. However, given that New Jersey venires are quite diverse, and given the

83

negative connotation of an “all-White jury,” the proportion of Whites on a jury

seemed a reasonable gauge for jury diversity.57

According to these data, there is a negative relationship between judges’ use

of cause challenges and the extent to which a jury contains non-Hispanic Whites:

the higher the proportion of cause challenges exercised in civil venires, the

lower the proportion of non-Hispanic Whites seated on a jury. Across the 95

trials, the correlation coefficient was -.29, which tested as statistically significant at

p < .01. However, when I divided the data by trial type, I found that the effect

existed significantly only among the 69 civil cases (r = -.30, p < .01); in the 26

criminal trials, the correlation was negative but non-significant (r = -.07, p < .71),

meaning that for criminal trials there was no significant relationship between the

57 Capturing jury diversity with these data is additionally challenging because a number

of people, including about 2% of jurors, did not provide a questionnaire that identified

their racial or Latino background (see Appendix A). In the absence of any other way to

estimate a person’s race, one must make various wholesale assumptions about the entire

category of people who did not provide a questionnaire. One can assume either that none

of them are White, all of them are White, or that the proportion of Whites who are

lacking questionnaires is similar to the proportion who did provide questionnaires. For

this section, I examined jury diversity by making all three assumptions and seeing

how/whether the different assumptions change results. As might be expected, the first

assumption underestimates the proportion of Whites on a jury, whereas the second

overestimates it. I therefore report results based on assuming that people with and without

missing data are comparable. Typically, this produced results that were in the middle of

results for the other two assumptions, and, further, as I describe in Appendix A, there are

sound reasons for assuming that those with and without surveys are comparable.

Therefore, the jury diversity measure described in this section reflects the proportion of

non-Hispanic Whites on a “jury” made up only of those who had questionnaire data.

84

proportion of cause challenges given and jury-composition outcomes. Nonetheless,

even the absence of an effect in criminal cases is noteworthy: Although use of

cause challenges does not improve jury diversity, neither does their use undermine

it.58

It is not clear why active use of challenges for cause should be associated

with more jury diversity and why the effect should emerge in civil rather than

criminal cases. Because I have only limited information about each trial in this

dataset, quite likely some third factor explains the relationship between cause

challenges and jury-composition outcomes. I conducted an analysis that controlled

for other characteristics of the venires, particularly the proportion of African

Americans, Latinos, and Asians in the venire – since, logically, more diverse juries

come from more diverse venires – as well as a control for the size of the venire.

The effect weakened but was nonetheless still negative.59

58 Consistent with results of Section V, in neither criminal nor civil cases did peremptory

challenges have any statistically significant relationship with jury diversity.

59 A multiple regression model permits several predictors to be entered at the same time

to predict the proportion of Whites on a jury, which allows the effect for one factor

(cause challenge use) to be assessed while controlling for other factors. According to this

model, on civil juries, size was unrelated to jury diversity, but the presence of higher

proportions of all three groups uniquely and negatively predicted the proportion of

Whites on the jury (the more diverse the venire, the lower the proportion of Whites on the

jury). Net of these factors, the percent of challenges for cause granted continued to be

negatively related to the proportion of Whites on the jury, but the effect fell to under

conventional levels of statistical significance (p < .08). In a similar regression on the

criminal trials, only the proportion of Blacks in the venire strongly and negatively

predicted the proportion of Whites on the jury (p < .0001).

85

Determining the basis for the effect would require a different study and a

better understanding of each county’s jury selection practices. It is noteworthy, for

example, that in Table VI.1, the counties with some of the lowest proportions of

cause challenge use in civil trials – Burlington, Camden, Gloucester, and Ocean –

are those counties reviewed in Section IV that were outliers in other ways: either

they were the lone areas that had consistent issues with minority representation

across both civil and criminal cases (Burlington and Ocean) or they consistently

seated no African American jurors in civil cases (Camden and Gloucester).

Although this offers an intriguing pattern, available data do not suggest that these

four counties fully drive the correlation. In most areas, the correlation was negative

(although there are two few trials within each county to reliably test for which are

significant); hence, the negative association between cause challenges and the

proportion of Whites on the jury was not restricted to just those four areas.

In sum, although I cannot fully explain the effect, I can conclude that judges

who more liberally grant challenges for cause should not fear that doing so

undermines jury diversity. Instead, actively dismissing people who raise

concerns about bias does not negatively affect diversity in criminal or civil cases.

C. Conclusions and Recommendation #6. Judges in New Jersey trials use

cause challenges with remarkable frequency. In criminal cases, challenges for

cause are the primary means of attrition from a venire, removing over half of all

86

venirepersons. In civil cases, cause challenges are second, just behind the “not

used” category, in ending jury service for people, and they account for just under

40% of the exits from jury service. Such substantial frequencies likely explain why

attorneys’ use of peremptory challenges and cause challenges are positively, not

negatively, related. Far from using more strikes to offset judges’ conservative use

of challenges, both tend to increase or decrease in tandem across cases, likely in

response to the level of concern about juror impartiality in the venire and other

factors. Juror race and ethnicity do not predict cause challenges: by and large,

judges grant cause challenges to the members of different racial/ethnic group at

comparable rates. Finally, active use of cause challenges does not harm jury

diversity in criminal cases, and particularly not in civil cases.

Recommendation #6. Judges in New Jersey courts should continue

their current practices in granting challenges for cause when they

deem them to be appropriate. The rate at which New Jersey judges

grant challenges for cause, while high, arguably allows attorneys to be

conservative in their use of peremptory challenges, and their use does

nothing to undermine jury diversity.

87

SECTION VII. THE SIZE OF JURY VENIRES

This report has uncovered a number of positive aspects of the jury selection

process in New Jersey: Most basically, New Jersey took the initiative to study its

practices by fielding the study that produced these data. Second, although New

Jersey resembles most jurisdictions by having pools of people appearing for jury

service that substantially underrepresent minorities (particularly African

Americans), and although there were pockets of concerning levels of minority

underrepresentation on juries, in the main I find that New Jersey does an admirable

job producing juries that mirror the venires. Third, despite statutes that grant

attorneys the opportunity to use a large number of peremptory challenges

(especially in criminal cases), the data showed a fairly constrained use of these

strikes, and only limited evidence that attorneys’ peremptories negatively affect the

representativeness of the final juries. Finally, New Jersey’s judges appear to

liberally grant challenges for cause, and no evidence suggests that these strikes

have a disparate impact on any racial/ethnic group. Judges’ approach to these

strikes may explain attorneys’ constrained use of challenges, and additionally,

judges’ challenges do nothing to undermine jury diversity.

A. The “not used” category. There remains one aspect of jury selection

practice in New Jersey that is less commendable. In criminal cases, the second-

most common way that people ended their voir dire experience was to be

88

labeled in the JMS database as “not used,” and this was the most common

way that people in civil cases ended their appearance at voir dire. According to

Table VI.1 above, in a few counties – Camden, Morris, and Ocean – the not-used

group accounted for more than 50% of the outcomes in venires, and in one of these

areas, “not used” made up more than 60% on venires.

The “right” number of people to call for voir dire will vary by a number of

factors and by local custom. However, other jurisdictions typically have more

modest-sized venires. For example, in a study of Federal courts I conducted, I

reported on interviews and other published figures regarding venire sizes:

A clerk from a Texas district court said that they typically bring in 45

people for voir dire in a routine, noncapital case; an attorney from a

California district said it was similar in her area, perhaps a little lower.

An attorney from a district near Chicago said that because their judges

tend to be generous with hardship excuses, they pull together panels

of about 60 . . . . Examples from a manual on federal selection

procedures indicate these are consistent with, if slightly larger than,

venires in some areas

(http://www.fjc.gov/public/pdf.nsf/lookup/jurselpro.pdf/$file/jurselpro

.pdf). These estimates are also somewhat higher than, but broadly

consistent with, venire sizes in [some] routine state courts (e.g., Rose

1999, reporting venires of between 30 and 40).60

Federal courts, the focus of the above quotation, differ from state courts in a

number of ways (e.g., judges tend to be highly restrictive about voir dire; trials are

60 Rose et al., supra, note 18 at 389.

89

rarer; and pools are comparatively less diverse). Yet Table VII.1 shows that venire

sizes in this study, particularly in criminal cases, were dramatically far from 50 –

60: venires in criminal trials averaged 144 people (range: 60 – 275).

Table VII.1: Average Venire Size, by County and Overall, for All Cases and By

Case Type.

County

N

Criminal

Trials

N Civil

Trials

Average

Venire

Size (All

Cases)

Average

Venire

Size

(Criminal)

Average

Venire

Size

(Civil)

Bergen 2 11 84.0 275.5 49.2

Burlington 2 6 41.6 61.5 35.0

Camden 1 7 74.8 194.0 57.7

Cumberland 2 0 88.5 88.5 --

Essex 5 7 91.0 130.8 62.6

Gloucester 1 2 92.3 175.0 51.0

Mercer 3 2 109.0 159.3 33.5

Middlesex 1 8 48.1 60.0 46.6

Monmouth 1 8 73.9 91.0 71.8

Morris 0 3 58.3 -- 58.3

Ocean 1 4 66.0 133.0 49.3

Passaic 3 6 112.9 226.0 56.3

Somerset 1 0 125.0 125.0 --

Union 3 5 68.6 104.7 47.0

Overall 26 69 78.0 144.3 53.0

The general picture the data paint is of venires that pull together more

jurors than is necessary to intelligently exercise challenges for cause, exercise

peremptory challenges, and seat a jury. The excess number of jurors who do

not fall into any of these categories – that is, who wind up being not used –

90

does nothing to enhance the diversity of final juries and risks making citizens

feel that their time is wasted.

After advocating in Section VI that judges should continue to be generous in

granting cause challenges, I do not argue here that New Jersey should drastically

reduce its jury sizes. The plethora of jurors in criminal venires likely helps to

account for why judges feel comfortable granting cause challenges. Indeed, in

criminal cases, the size of the venire is strongly and positively associated with the

rate of cause challenges (r = .54, p < .01). In civil cases, in which the range of

venire sizes is more attenuated, the association is less strong and only marginally

significant, but still positive (r = .21, p < .09).

At the same time, there are negative consequences associated with having

large numbers of people fall into the “not used” category. Although those “not

used” may report that they are happy to avoid having to serve further, research

suggests that people resent courts’ wasting their time.61 Beyond perception, many

of those who appear at court have made actual sacrifices to appear, in terms of both

time and money. Anything that can lessen the likelihood that people must take time

to appear for service if they are not, in fact, needed will tend to boost the

61 See, e.g., William R. Pabst, Jr., G. Thomas Munsterman, and Chester A. Mount. The

Myth of the Unwilling Juror, 60 Judicature 164 (1976); Rose, supra, note 46; see

generally, Marika Litras, and John R. Golman, A Comparative Study of Juror Utilization

in U.S. District Courts, 3 JOURNAL OF EMPIRICAL LEGAL STUDIES 99 (2006).

91

legitimacy of New Jersey courts in people’s mind. Lowering the rate of the “not

used” category would further that aim.

It is perhaps axiomatic to assume that bringing in more people to a venire

and creating larger groups will enhance jury diversity. However, in both civil and

criminal cases, the absolute size of a venire was uncorrelated with the diversity of

the final panel (r = -.01 in both civil and criminal cases). Therefore, large venires,

by themselves, do not accomplish diversity goals. Instead, the best correlate of a

diverse jury is a diverse venire. As described above (see Footnote 49), in

criminal cases the sole significant predictor of higher levels of non-Whites on a

jury in criminal cases was the percentage of African Americans in the venire. In

civil cases, increases in the proportions of Blacks, Latinos, and Asian Americans

all significantly predicted greater jury diversity. This is another reason why New

Jersey should do all it can to improve the representativeness of those who appear at

court, particularly African Americans, but certainly all groups (see Section III and

Recommendation #3).

But even with the current groups of people who come to jury service, large

numbers of “not used” persons indicate that the resources of both courts and people

are being squandered. New Jersey should therefore critically examine why so

many people called to a voir dire wind up “not used.” Although there will always

be uncertainty over the right number of people to call to a courtroom for a specific

92

case, in civil cases close to 40% of individuals called up are not used; with average

venire size in the mid-fifties, this amounts to over 20 people who go to voir dire

unnecessarily in a typical case. In criminal cases, the proportion of not used is

smaller – at just over one-quarter of venires – but given the large venire sizes, the

expected number of people who will be not used is larger in absolute terms (e.g.,

using the average values in Table VII.1, .267 x 143 = 40).

Although I do not have data to fully explain the large venire sizes, two

sources are likely. Particularly in criminal cases, judges may plan for all

peremptory challenges being used; as noted, this does not typically occur, but it

remains a risk. If planning for attorneys’ using large numbers of peremptory

challenges contributes to the large panel sizes described in Table VII.1, then

this is a strong, empirically-supported reason to consider reducing the

number of challenges allotted, particularly in criminal cases. This rationale is

stronger, in my assessment, than the peremptory’s potential impact on jury

diversity, which, as Section V showed, the data support only weakly. The data

make clear that large numbers of people called up for cases are, literally, not

needed; to the extent that peremptory challenge allotments account for this, then

that allotment should be reconsidered.

Second, judges’ willingness to support challenges for cause may depend

upon having large panels to avoid “running out” of prospective jurors. If so, it may

93

be necessary to convene judges statewide to discuss explicitly what amount of

“surplus” is necessary to permit them to be comfortable granting cause

challenges they deem appropriate. The latter would achieve the goal of lowering

venire sizes while preserving a practice that appears to constrain peremptory

challenge use and that does not harm to jury diversity. Thus, my final

recommendation to New Jersey is to focus scrutiny on the size of venires.

Recommendation #7: New Jersey should determine ways to reduce

the number of people who are called to voir dire only to be “not

used.” Possible mechanisms include reducing the number of

peremptory challenges, particularly in criminal trials, or convening

judges to consider by how much venire panels might be reduced while

still allowing judges to be comfortable during cause challenge

determinations.

94

APPENDIX A: HANDLING MISSING QUESTIONNAIRE RESPONSES

AND MISSING QUESTIONNAIRES.

The pattern that appears in Table II.1 is clear: Although only about 3% of

people who filled out a questionnaire failed to indicate their race, that figure was

nearly four times as high for the Latino/a ethnicity question. This pattern is typical

whenever questionnaires are designed the way New Jersey’s was, which asked first

about race and then about Latino/a ethnicity.62 Although the Census Bureau makes

a distinction between race and Hispanic ethnicity – recognizing that people might

be White and Hispanic (e.g., they are from Spain) or Black and Hispanic (e.g., they

are from the Dominican Republic) – a substantial number of people treat Latino/a

ethnicity as a racial category.63 Hence, when people first report their race and are

then asked about their Hispanic/Latino ethnicity, some tend to mistakenly view the

second question as redundant with the first ( i.e., they have already reported their

race), which leads to higher levels of missing data for the Latino/a question than

the race question. For this reason, the Census Bureau re-ordered how they asked

62 See, e.g., Mary R. Rose, & Jeffrey B. Abramson, Data, Race, and the Courts: Some

Lessons on Empiricism from Jury Representation Cases, 2011MICH. ST. L. REV. 911

(2011).

63 See, e.g., Ann Morning, Keywords: Race, 4 CONTEXTS 44, 44 (2005).

95

about these two issues on their questionnaires, first asking about Hispanic/Latino

ethnicity and then asking about race.64

These non-response missing values compound the problem of data missing

because of response rates to the questionnaire at all, but regardless of source, all

forms of missing data are challenging to address in studies such as this one. There

are, for example, no other pieces of information in the file that might permit one to

make an educated guess about the person’s likely race or ethnicity, at least for

purposes of generating informed estimates of a group’s likely racial/ethnic

composition. For example, in communities with high levels of racial segregation in

housing, a street address can be a fairly reliable marker of an individual’s likely

race or ethnicity. For Latino/a ethnicity, sometimes a person’s name may signal

their likely background.

Absent these types of additional information, the analyst must instead make

a blanket decision about whether to include missing cases in analyses or not. If the

analyst excludes all missing data, she does so on the assumption that retained cases

and the group of people with missing values are functionally the same (social

scientists call this data that is “missing at random”) – that is, those with non-

64 E.g., Nancy Bates, Elizabeth A. Martin, Theresa J. DeMaio & Manuel de la Puente,

Questionnaire Effects on Measurement of Race and Spanish Origin, 11 J. OFFICIAL

STAT. 433, 433 (1995).

96

missing data are a representative sub-sample of all those in the study, and

analyzing only their data will yield the same results as the full set. Whether it is

plausible to assume that data are missing at random depends on the source of the

missing data. In the New Jersey study, there were multiple sources of missing data,

and the plausibility of randomness differs across these sources.

A. Skipping the Hispanic/Latino or the race question. For people who filled

out a survey but skipped the Hispanic/Latino question, the notion that the data are

missing at random is not empirically supported and is therefore implausible. For

example, of all those filling out surveys, 12% self-described themselves as

“African American”; among those with missing data on the Latino/a ethnicity

question, fully 24% were African American; likewise, Asians were 9% of all

survey respondents but 13% of those missing on the Latino question. Both of these

results signal that those who skipped the Latino/a question are not simply a random

subset – i.e., the same as – the group who answered it. As noted above, in all

likelihood they believed they had already answered the race question. In this

situation, if an analyst were to simply remove all missing data due to skipping this

question, those removed are likely to be disproportionately non-Latino. One has

lowered the overall denominator that generates the proportions of each group (e.g.,

the percent in the sample who is Latino/a), and has removed from that denominator

more non-Latinos than Latinos, artificially inflating the number of Latinos in the

97

sample. That is, it will appear that Latinos are better represented in jury pools than

they actually are. For purposes of examining racial representativeness, a more

careful and conservative approach is to retain all missing cases and code them all

as non-Hispanic/Latino, which is the approach I took in this study.65

Likewise, it does not appear that people who skipped the race question (n =

302 people) are a random subset of the whole sample. Instead, fully 57% of those

who skipped the race question (n = 171 people) answered that they were

Hispanic/Latino. Plausibly, these respondents did not “see” their group in the race

question and instead indicated it in the Hispanic/Latino question. Of the remaining

non-respondents, most (n = 114 people) failed to answer both the race and the

ethnicity question; only 17 people skipped the race question and also said they

were not Latino. Therefore, for cases that are missing on race because a respondent

skipped the race question, the most conservative approach is to code that person

into the “Other” group. This is because “Other” is the most common category

65 I recognize that this may tend to underreport how many Latino/a persons appeared at

the courthouses, since undoubtedly some of the people who skipped the question were

Latino. But because New Jersey seeks to understand the racial and ethnic

representativeness of its pools, my approach of retaining missing cases and assuming

they are non-Latino offers the most conservative estimate.

98

selected by Latinos (34%).66 Of those who selected this category on the

questionnaire, 90% were Latino.

B. Missing whole surveys. In contrast to those instances in which someone

skipped a particular question, the missing-at-random assumption is far more

plausible when data are missing because someone did not take the survey (or took

it, but it was not collected by the court). Among those who made it to voir dire, for

example, there is no statistically significant relationship between lacking a survey

and type of outcome.67 For this reason, I omit those with missing surveys when

conducting analysis of the composition of jury pools/venires. In other words, the

absolute and comparative disparities described in Section III are based only on

those individuals who provided the court with a survey. For statistical analyses of

outcomes among the group who made it to voir dire (see Section IV), I ran models

both with and without those who lacked a survey, and results did not differ.

C. Handling multiple answers on race. A final complexity in measurement

of race in this study stems from the fact that the questionnaire instructed people

66 The next-most commonly selected category among Latinos is “White” (32%), which is

one reason to remove people who mark both “White” and “Latino” from the “White”

category when making comparisons between Latinos and Whites, i.e., to avoid having

people in both categories. When I have done this in the report, I refer to Whites as “non-

Hispanic White” or “White/non-Hispanic.”

67 Chi-square test of association = 8.48, df = 4, N = 7,407 (p < .08). Across most

categories of results, about 12 – 14% had missing surveys.

99

that they could mark more than one category for race (the parenthetical next to the

word “Race” said: “check all applicable categories”). This approach has the

advantage of being highly inclusive and respectful of the fact that many people in

the United States have a rich and complex heritage that includes ancestors – or

even just birth parents – from different racial groups. Indeed, the Census Bureau

allows people to describe themselves through multiple categories.68

For the types of analyses undertaken here, however, this more inclusive

approach has some notable limitations. According to U.S. Supreme Court

jurisprudence, proving underrepresentation first requires that a party prove that

those underrepresented on the jury form a “cognizable group.”69 Such groups are

typically understood to be made up of a single racial identity that reflects a group

with a shared history and set of experiences (e.g., African Americans, Latinos,

Asian Americans, Native Americans, and women). I expect it would be

challenging to show that a group described as “Multiracial” constitutes a

cognizable group, at least without additional information about which groups make

up this category specifically.70 Although this report is not intended to support

68 See, e.g., https://2020census.gov/en/about-questions/2020-census-questions-race.html.

69 Duren v. Missouri, 439 U.S. 357 (1979).

70 I know of no attempts to pursue a Fair Cross Section claim on the basis of the

underrepresentation of “Multiracial” people, and I suspect it would be hard to win on any

such claim. By way of analogy, empirical research on discrimination claims shows that

100

litigation on representation, the concept of a cognizable category has resonance for

a study of representativeness.

Second, even though multiracial persons may identify with multiple racial

backgrounds, they may also present to other people– including to attorneys, judges,

or fellow jurors interacting with and making assessments about them – as members

of but one category in their background. A clear example is Tiger Woods, who is

both Asian and African American but is typically identified as “the first African

American” to obtain a particular milestone in golf. 71 Likewise former President

Barak Obama is often described as the first African American to become president,

even though he also has a White parent. A question that permits people to put

themselves into multiple racial categories elides the issue of whether or not the

individual believes his or her own lived experience is closer to one racial identity

more than another.

Finally, although the number of people who identify as “Multiracial” may be

growing in the U.S., they typically form a small proportion of people in most

cases in which people claim discrimination on the basis of multiple identities – called

“intersectionality” – lowers the likelihood that the plaintiff will win. See: Rachel Kahn

Best, Lauren B. Edelman, Linda Hamilton Krieger, and Scott R. Eliason. Multiple

Disadvantages: An Empirical Test of Intersectionality Theory in EEO Litigation, 45 LAW

& SOCIETY REVIEW 991 (2011).

71 See, e.g., https://www.pga.com/story/timeline-of-african-american-achievements-in-

golf.

101

studies. In this study, just 1.5% of people selected “Multiracial” (although others

selected it in combination with other designations). In conducting statistical

analyses, small-sized categories are problematic because one cannot form reliable,

plausible estimates on a group that is comparatively much smaller than other

groups. In that situation, members of the group are either combined with some

other group (e.g., “Multiracial/Other,” which I did in this study), or in the worst

case, the category must be omitted all together – wholly undermining the intended

purpose of being more inclusive when asking about race. Asking people to indicate

the “best describes” racial category would limit the number of people who must be

recoded into/combined with another group to increase statistical power, or limit the

number who are omitted from analyses.

As a practical matter, I had to decide how to code people who checked more

than one box on the race question for purposes of analyzing the racial

representativeness of New Jersey’s jury pools, venires, and juries. In one way, this

was not a particularly challenging task: just 5% of respondents to the survey

marked more than one category, whereas 95% singly marked either Black, Asian,

Multiracial, Other, or White. However, given that 15% of people in the dataset

were entirely missing on race (due to non-response) and another nearly 3% opted

not to answer the race question at all, I did not want to lose more people from

single, cognizable racial categories than was necessary.

102

In New Jersey, the most populous racial categories are White, Asian, and

African American. Given its long history of exclusion from jury participation in

the United States, the latter group seemed especially important to capture with as

much reasonable accuracy as possible, and I therefore prioritized identifying those

individuals who had Black ancestry. For individuals who marked more than one

category for race, I did the following recodes:

*A person was coded as “Black/African American” if that person marked

either “Black” (alone) or marked “Black” in conjunction with one other

racial group (e.g., “Black/Asian” or “Black/White). A total of 33 people who

marked two categories, one of which was “Black” were recoded as “Black.”

If someone marked three or more racial categories that included “Black,”

that person was coded as “Multiracial.”

*A person was coded as “Asian” if that person marked either “Asian”

(alone) or marked “Asian” in conjunction with one other racial group

besides “Black/Asian.” A total of 17 people who marked “Asian” and one

other category (exclusive of “Black) were recoded as “Asian.” If someone

marked three or more racial categories that included “Asian,” that person

was coded as “Multiracial.”

*A person was coded as “White” if that person marked either “White”

(alone) or marked “White” in conjunction with one other racial group

besides “Black/White” or “Asian/White.” A total of 54 people who marked

“White” and one other category (exclusive of “Black” or “Asian”) were

recoded as “White.” If someone marked three or more racial categories that

included “White,” that person was coded as “Multiracial.”

D. Conclusion and Recommendations #2a, #2b, and #2c. As I noted in

Recommendation #1, New Jersey should develop a plan for more routinely

measuring and keeping records about the demographic. The experience of this

103

study shows that, whether they develop this system or embark on additional stand-

alone studies like the current effort, a few principles of questionnaire design will

lessen the likelihood of missing data and will allow results to be consistent with

existing jurisprudence on jury representation. The aims are (a) to lower the

likelihood that people do not answer questions; (b) give people the opportunity to

indicate the racial group that best describes how they identify themselves; and (c)

remove confusion about whether or not they should answer the question about

Latino/Hispanic ethnicity. Therefore, I make three recommendations about

questionnaire design in any future systems of measuring race/ethnicity in jury

pools:

Recommendation #2a: Because of the policy and legal importance of

understanding the demographic patterns in a jury selection system,

people should not be specifically invited to consider questions about

race, ethnicity, or gender to be voluntary. This will lower the

likelihood that people fail to turn in a survey at all or that they skip

questions (e.g., by indicating that the question is optional). New

Jersey may take a position that such questions should be voluntary. If

so, court personnel can treat the survey that way by not admonishing

people who fail to turn one in, not returning questionnaires to people

who have skipped questions, and by programming any online

questionnaires so that people can skip a question. But neither the text

of the questionnaire itself nor the race/ethnicity questions should

invite people to fail to answer. Instead, people should be informed

about the reasons why the questions are necessary to ask, with a brief

explanation accompanying any questions about race, ethnicity or

gender.

Recommendation #2b: Because the concept of a “cognizable group”

is one requirement for proving that underrepresentation has occurred

in the jury summoning, qualification, or selection process, questions

104

about race should not invite people to select multiple categories.

Instead respondents should be asked to select the category that “best

describes” their race. Options appearing on the form can include a

“Multiracial” category so that people who feel that this designation

best describes them can select it. However, a “best-describes” design

would increase the likelihood of identifying those people who may be

multiracial but who tend to identify with and experience their race

primarily through one aspect of their background more than another; it

also reduces instances in which people are part of a group that is too

small in size to reliably analyze.

Recommendation #2c: To minimize the tendency to skip a question

about Latino ethnicity whenever it follows rather than precedes a

question about race, respondents should be asked about whether or not

they are Latino/a before they are asked to identify the racial group that

best describes them.

105

APPENDIX B: STATISTICAL TESTING OF UNDERREPRESENTATION

ON JURIES

As described in Section III, scholars of representativeness prefer the

comparative disparity test as the best way to convey whether underrepresentation is

“not fair and reasonable,” and throughout the report, I adopt a threshold of at least

a 25% comparative disparity to call underrepresentation “substantial” or

“concerning.” A wholly other way to look at underrepresentation is to ask whether

the likelihood that members of one group will become jurors differs statistically

from the likelihood that exists for members of another group. Here, the question is

whether it is reasonable to assume that the likelihood of becoming a juror is the

same for different groups. Scholars call this approach testing against the “null

hypothesis,” or testing against the hypothesis that the difference in likelihoods for

the different groups is zero.

A. The meaning of a p-value. Unlike a comparative disparity, in which

larger values get more notice, in statistical testing a low probability is the relevant

threshold; that threshold stems from a commonly-accepted statistical metric called

a “p-value” (or, probability-value). In most social science disciplines, a probability

is considered “low” if it is lower than 1 in 20, or p < .05. The meaning of this p-

value and this likelihood is quite technical and specific, so it bears describing here.

106

One way to think about a p-value is to imagine a hypothetical world that

contains a large venire of, for example (and for simplicity), equal numbers of

African American and White members. Further imagine in this hypothetical that

selection into a sub-group (e.g., to become jurors) is, in fact, exactly equal for

every member; it is equal because samples are drawn completely randomly – that

is, this hypothetical is nothing like actual jury selection in the real world. With

random sampling, every member of the venire has exactly the same probability of

being selected into the subgroup called “jurors”; the difference in the true

likelihood of selection would be, literally, zero.

However, every time a sample is drawn, this sub-sample will differ some

from the larger population; this is called “sampling error.” Usually, given random

sampling, the differences between the population and sampled sub-group will be

small, but sometimes the difference, by chance, will be quite substantial. For

example, by chance, twice as many White people as African Americans could be

selected for the juror-group, even though, in the hypothetical, the prevalence of the

two groups in the population is exactly the same and selection is random. Although

such discrepancies could happen, truly large discrepancies like this one – with

twice as many Whites and Blacks on a jury when there should have been equal

numbers of each – will be unusual and rare. Indeed, the bigger the discrepancy

between the profile of the sampled group and the profile of the population (again,

107

in this hypothetical null world of random sampling), the lower the probability that

this could occur if there should have been no difference.

Thus, when conducting statistical testing, one is basically asking: “What is

probability that I got this big of a difference in jury participation between these

groups when, given the null hypothesis, their likelihood of participation should

have been exactly equal?” P-values quantify the precise probability of a given

outcome under the assumption of a null hypothesis, i.e., that there should have

been no difference in outcomes. If that likelihood is small – typically set at a

threshold of less than one in twenty, or p < .05 – then we term that difference

“statistically significant.” P-values depend on many factors, including the size of

groups that were sampled/observed in a study, the prevalence of the outcome being

predicted, and the properties of the distribution of different statistics under the null

hypothesis.

B. Choosing the appropriate statistical test. There are various ways to test

whether the likelihood of becoming a juror differs across groups, and which

approach to take depends on the structure of one’s data. In this study, for example,

a statistical test must account for the fact that each observation in the dataset is part

of a trial, and the basic likelihood of becoming a juror differs across each trial (e.g.,

the venire sizes are different, as are the jury sizes). I therefore used a mixed-model

logistic regression, with trial number accounting for the fact that observations are

108

clustered together by trial. For tests described in this section, becoming a juror (or

not) was the outcome variable, and the different racial groups served as predictors

of this outcome. I coded groups as “Black,” “Asian,” and “Other/Multiracial” (see

Appendix A for more information on putting people into racial groups); “White”

served as the reference category - that is, the question was always whether the

likelihood of becoming a juror for one group (e.g., “Blacks”) differed from the

likelihood that White venire-members would become jurors. I ran separate models

for race, and then for Latino ethnicity (with Latinos compared to non-Latinos).

Finally, I also tested a model that interacted race and gender, in case, for example,

Black Males have a different probability of selection compared with White Males.

C. Results for predicting jury participation. The statistical tests revealed no

instance of a statistically significant level of underrepresentation in the New Jersey

trials that were part of this study. Exactly one comparison produced a near-

significant effect: Asians were slightly less likely to be selected as a juror

compared to Whites, but the p-value was just shy of conventional significance, p <

.07. Further, that result is sensitive to the counties studied. When I restricted the

test to only those counties in which Asians make up more than 5% of the venires,

the effect went away entirely (p < .23).

To ensure that counties with strong levels of racial representation were not

off-setting counties with comparatively poorer levels, I re-ran the models by

109

county. This did not change results. Results were also non-significant when I

included in the model variables designed to test whether any race effects depended

upon the gender of a person (e.g., African American men are disproportionately

underrepresented on juries, compared to other groups); those tests were also non-

significant. Turning to tests of whether Latinos are underrepresented on these

juries, there was no statistically significant effect in the overall model, no

statistically significant underrepresentation in models examined by county, and no

evidence that any effects for being Latino depend upon the gender of a person.

Although it is certainly impressive that rates of participation for Blacks,

Latinos, Asians, and others are not statistically lower than rates for non-Hispanic

Whites, there exists one serious limitation of using statistical significance as a lens

on these data. As Section VII of this report details, New Jersey uses quite large-

sized jury pools to select jurors. This means that for everyone, being selected for a

jury is a low-probability event: about 11% of people in New Jersey jury pools were

selected for a jury. When the likelihood of being selected is low for everyone, then

only the very largest disparities will test as statistically significant. For example,

one significant result did emerge in all the tests I ran: Latinos tested as

significantly overrepresented on juries in Morris County. However and notably,

their prevalence on juries was over three times as high (28.6%) as their prevalence

in the venire (8.9%); for differences smaller than this type of magnitude,

110

particularly when the group’s prevalence on the venire is even smaller than this

example, the null hypothesis – that differences are due to sampling error – remains

plausible. In short, statistical significance is but one way to view levels of

underrepresentation, and particularly in analyses of participation on a jury, a

disparity must be extremely large to be “significant.” Few disparities in the New

Jersey data were that large in size.